data_608_final_project

OVERVIEW

Fears of the coronavirus crashed the stock market back in February, precisely beginning on February 24, 2020. The pandemic sent a shockwave to the global market and it still continuously wreaked havoc to humanity. The fears spread quickly and globally, e.g. over 70% of the world population was under lockdown at some point in March. Recently in May, the US unemployment rate has climbed to historical high since the Great Depression, and the whole country is still not ready for reopen (as of this moment the report is generated on 2020-05-13 18:42:36). Nevertheless, the US stock market has remarkably bounced back since late March, early April and continuously rising in May. The inspiration of this final project is come from the decoupling of the stock market and reality. Can we predict and visualize any upward or downward trend of any stock during this significant crisis in our time?

(I) SET UP

We first load a list of packages in the session using pacman::p_load. We need to register for an API from https://www.alphavantage.co/ in order to get real time data using the quantmod package. It’s free to use this service as long as there’s not more than 500 calls a day.

# load packages
packages <- c('broom', 'caret', 'doParallel', 'e1071', 'foreach', 'glue', 'gridExtra', 'Hmisc', 'htmlwidgets', 'InformationValue', 'kableExtra', 'parallel', 'plotly', 'quantmod', 'sqldf', 'tictoc', 'tidyverse', 'TTR', 'wrapr')
pacman::p_load(char = packages)

# set environment
readRenviron("~/I/config/.env")

# set project home directory
PROJECT_HOME_DIRECTORY <- Sys.getenv("PROJECT_HOME_DIRECTORY")
setwd(PROJECT_HOME_DIRECTORY)

# get real time data if necessary
REAL_TIME = TRUE
ALPHA_VANTAGE_API <- Sys.getenv("ALPHA_VANTAGE_API")

(II) GET DATA

For illustrative purpose, we pick only one stock to look at in this final project (but the script is built to handle multiple stocks at the same time). We decide to choose a very popular stock to look at, i.e. Apple Inc (AAPL). When downloading historical data for multiple stocks, it’s wise to use multicores on one’s machine and create a parallel processing to speed up the data retrieval processing.

# stock symbol
stock <- c("AAPL")

# detect, use multicores
numCores <- parallel::detectCores()

# create a simple cluster on the local machine using all available threads
cl <- parallel::makeCluster(detectCores(), methods = FALSE)

# register our cluster
doParallel::registerDoParallel(cl)

# loop through a list of stock tickers - super fast! 5x faster than the traditional for-loop approach!!
symbols <- foreach::foreach(i = 1:length(stock), .errorhandling = 'remove') %dopar% { quantmod::getSymbols(stock[i]) } %>% unlist  # change .errorhandling = 'pass' to see error

# return a single list of xts objects from the valid symbols
if(REAL_TIME){
        
        xtsList <- vector(mode = "list", length = length(symbols))
        
        for(i in 1:length(symbols)){
                
                xtsList[[i]] <- quantmod::getSymbols(
                        
                        symbols[i], 
                        env = NULL,  # set env = NULL and that is equivalent to auto.assign = FALSE
                        src = "av",
                        periodicity = "daily", 
                        output.size = "full", 
                        adjusted = TRUE,
                        api.key = ALPHA_VANTAGE_API
                        
                )
                
                Sys.sleep(12)
                
        }
        
} else {
        
        xtsList <- foreach::foreach(i = 1:length(symbols)) %dopar% { quantmod::getSymbols(symbols[i], env = NULL, adjusted = TRUE) }  # set env = NULL and that is equivalent to auto.assign = FALSE        
        
}

# set names for xtsList
names(xtsList) <- symbols

(III) BASIC TECHNICAL ANALYSIS

Before diving into our prediction model, let’s take a quick look of the time series data from “01-02-2019” to “05-13-2020”. There are two obvious observations from the chart, i.e. 1) there’s a steady upward trend presented in the data, and 2) there’s a significant and rapid down turn (caused by the pandemic) starting on “02-24-2020” until a bounce-back on “03-23-2020”. The chart includes multiple technical indicators, e.g. Bollinger Bands, Width and %B Indicator, Average Directional Movement Index (ADX), Moving Average Convergence Divergence (MACD), Relative Strength Index (RSI), Commodity Channel Index (CCI), Chaikin Money Flow (CMF), and Trading Volume.

(IV) DATA TRANSFORMATION, BUILD MODELS

There are four simple concepts that we need to first go over, i.e. i) nday (or tday as they are interchangable throughout the script), ii) lday, iii) trend, and iv) close_percentile_threshold.

nday is referred to the number of future trading day(s). For example, we want to use today (t) closing price to predict tomorrow (t plus 1) closing price, and we want to know whether the difference (closing tomorrow - closing today) / closing today would signal an upward or downward trend. Therefore, we actually look at two trends, and we simply refer them as “bullish” and “bearish”.

How should we set up a threshold for triggering a “bullish” or “bearish” signal? We look at the change in percentage between today and future values. Say, we want to make a prediction for tomorrow (or the next trading window), we first calculate the ratio of (closing tomorrow - closing today) / closing today, and then we find the distribution and look up the percentile (see below density curves for better illustration). Anything that is roughly above 1 SD is considered to be an upward trend (“bullish”) signal, whereas anything that is roughly below 1 SD (“bearish”) is considered to be a downward trend signal. We decide to choose roughly “plus or minus one SD” as our close_percentile_threshold, i.e. above 85th percentile / below 15th perentile. Thus our target of prediction is always a moving target based on market volatility and the gap between current and future trading windows. We decide to go with a very liberal approach (roughly plus or minus 1 SD) because we do not want to miss any signal. We want to avoid any type II error.

In addition, we need to transform our data by getting values from the past to see if that would predict the future. We use “5 ndays” and “5 ldays” in this final project. The 5 ndays are 1, 2, 3, 4, and 5, whereas the ldays are 1, 5, 10, 15, and 20. What that means is that for each n day, we use 5 ldays and each lday would make an individual prediction for the nday. Subsequently, we would create a consensus out of the 5 models. For example, let’s say we use trading volume to predict closing price. We would transform our data and get the trading volume of today minus 1 (previous trading day), today minus 5 (5 days ago), 10 days ago, 15 days ago, and 20 days ago. Each set of data would make a prediction of today plus 1 (next trading day), plus 2, 3, and so forth. We need to find a consensus out of these five models and that’s our signal for trend.

We write a custom function to transform and return values plus multiple indicators (such as RSI, MACD) from the time series object that we pull from quantmod. Please see the appendix for detail.

After data transformation, we build separate classification models to identify whether future trading day will close at higher or lower end than the present day. Our target is actually a threshold, i.e. whether it is above or below 1 SD and we use it as our predicted or dependent variable (“Yes” or “No”).

We build our classification/prediction model using different algorithms, i.e. glm (generalized linear model), nb (naive bayes) and gbm (gradient boosting machine). We build a base layer (using glm, nb and gbm) to predict the target from the train set. Second, we build a top layer (again using glm, nb and gbm) to predict the target based on the outcomes (OOF or Out-Of-Fold prediction) of our base layer. In other words, we make prediction (of the target variable) based on the predictions of our base models. Finally, we simply average the outcome probabilities of our top layer to get our final probability and decision using 0.5 as cut-off.

Instead of using entire historical record, let’s just focus on the past 1000 trading days. We split the data by 80/20 for our train and test sets respectively. Finally, we use the past 90 trading days as our evaluation set for accuracy.

Our focus is to see whether our data transformation (using ldays 1, 5, 10, 15, 20 to predict for ndays 1, 2, 3, 4, 5) and ensembling different algorithms would return any signal about stock price movement.

(V) RESULTS

Here we show a snippet of the original data, then the same piece of data after transformation with additional features, and subsequently, we scale and impute missing data and split it into train and test sets respectively.

Down below is a summary of the model output. We care mainly about the model of “final - averaging” and that’s the result of our top layer ensembling our models. We apply that specific model to our evaluation set using the most recent 90 trading days (including today closing price on 2020-05-13) to generate prediction. The target field represents the actual, whereas the pred_final represents our prediction.

original data

AAPL.Open	AAPL.High	AAPL.Low	AAPL.Close	AAPL.Volume	AAPL.Adjusted
300.46	303.240	298.87	300.63	35583438	299.8372
303.22	305.170	301.97	303.74	28803764	302.9390
305.64	310.350	304.29	310.13	33511985	310.1300
308.10	317.050	307.24	315.01	36486561	315.0100
317.83	319.688	310.91	311.41	40575263	311.4100
312.15	315.950	303.21	307.65	49979647	307.6500

after transformation

	trend	symbol	l_period	n_period	partition	date	year	quarter	month	weekday	day	open	high	low	close	volume	adjusted	rsi_close	cci_close	macd_close	signal_close	macd_close_diff	op_l	hi_l	lo_l	cl_l	vol_l	adj_l	op_n	hi_n	lo_n	cl_n	vol_n	adj_n	cl_op_diff	hi_lo_diff	op_l_diff	hi_l_diff	lo_l_diff	cl_l_diff	vol_l_diff	adj_l_diff	op_n_diff	hi_n_diff	lo_n_diff	cl_n_diff	vol_n_diff	adj_n_diff	op_l_percent_diff	hi_l_percent_diff	lo_l_percent_diff	cl_l_percent_diff	vol_l_percent_diff	cl_op_diff_by_op	hi_lo_diff_by_op	cl_op_diff_by_cl_l	hi_lo_diff_by_cl_l	cl_cl_l_ratio	op_cl_l_ratio	vol_cl_l_ratio	vol_op_l_ratio	vol_diff_ratio	sma_5	sma_10	sma_15	target
996	bearish	AAPL	20	5	5	2020-05-06	2020	2	5	Wednesday	6	300.46	303.240	298.87	300.63	35583438	299.8372	62.79564	153.90974	2.729758	1.968072	0.7616867	270.80	271.70	259.00	259.43	50721831	258.7459	312.15	315.95	303.21	307.65	49979647	307.65	0.17	4.370	29.66	31.540	39.87	41.20	-15138393	41.0913	11.69	12.71	4.34	7.02	14396209	7.8128	0.1095273	0.1160839	0.1539382	0.1588097	-0.2984591	0.0005658	0.0145444	0.0006553	0.0168446	1.158810	1.158154	-0.2575566	-0.2577024	-0.9509734	294.844	288.170	284.8393	No
997	bearish	AAPL	20	5	4	2020-05-07	2020	2	5	Thursday	7	303.22	305.170	301.97	303.74	28803764	302.9390	64.18093	162.03047	2.958091	2.166075	0.7920153	262.74	267.37	261.23	266.07	42223821	265.3684	NA	NA	NA	NA	NA	NA	0.52	3.200	40.48	37.800	40.74	37.67	-13420057	37.5706	NA	NA	NA	NA	NA	NA	0.1540687	0.1413771	0.1559545	0.1415793	-0.3178314	0.0017149	0.0105534	0.0019544	0.0120269	1.141579	1.139625	-0.2784138	-0.2788913	NA	296.832	291.041	285.9760	NA
998	bearish	AAPL	20	5	3	2020-05-08	2020	2	5	Friday	8	305.64	310.350	304.29	310.13	33511985	310.1300	66.90741	187.14544	3.273022	2.387465	0.8855568	268.70	270.07	264.70	267.99	40529123	267.2833	NA	NA	NA	NA	NA	NA	4.49	6.060	36.94	40.280	39.59	42.14	-7017138	42.8467	NA	NA	NA	NA	NA	NA	0.1374767	0.1491465	0.1495655	0.1572447	-0.1731382	0.0146905	0.0198272	0.0167544	0.0226128	1.157245	1.140490	-0.1496124	-0.1518103	NA	301.044	293.757	287.7980	NA
999	bearish	AAPL	20	5	2	2020-05-11	2020	2	5	Monday	11	308.10	317.050	307.24	315.01	36486561	315.0100	68.85704	187.12734	3.607397	2.631451	0.9759463	268.31	273.70	265.83	273.25	32755731	272.5294	NA	NA	NA	NA	NA	NA	6.91	9.810	39.79	43.350	41.41	41.76	3730830	42.4806	NA	NA	NA	NA	NA	NA	0.1482986	0.1583851	0.1557762	0.1528271	0.1138985	0.0224278	0.0318403	0.0252882	0.0359012	1.152827	1.127539	0.0987993	0.1010152	NA	305.414	296.941	290.3367	NA
1000	bearish	AAPL	20	5	1	2020-05-12	2020	2	5	Tuesday	12	317.83	319.688	310.91	311.41	40575263	311.4100	65.77831	134.84359	3.724404	2.850042	0.8743619	280.00	288.25	278.05	287.05	48748672	286.2930	NA	NA	NA	NA	NA	NA	-6.42	8.778	37.83	31.438	32.86	24.36	-8173409	25.1170	NA	NA	NA	NA	NA	NA	0.1351071	0.1090650	0.1181802	0.0848633	-0.1676642	-0.0201995	0.0276185	-0.0223654	0.0305800	1.084863	1.107229	-0.1545487	-0.1514269	NA	308.184	300.224	293.2060	NA
1001	bearish	AAPL	20	5	0	2020-05-13	2020	2	5	Wednesday	13	312.15	315.950	303.21	307.65	49979647	307.6500	62.62864	97.34025	3.669459	3.013925	0.6555334	282.40	286.33	280.63	284.43	32788641	283.6799	NA	NA	NA	NA	NA	NA	-4.50	12.740	29.75	29.620	22.58	23.22	17191006	23.9701	NA	NA	NA	NA	NA	NA	0.1053470	0.1034471	0.0804618	0.0816370	0.5242976	-0.0144161	0.0408137	-0.0158211	0.0447913	1.081637	1.097458	0.4847260	0.4777382	NA	309.588	302.216	295.3093	NA

train set

	partition	date	rsi_close	op_cl_l_ratio	hi_l_percent_diff	lo_l_percent_diff	op_l_percent_diff	cl_l_percent_diff	cl_cl_l_ratio	cci_close	macd_close	vol_n_diff	sma_15	hi_lo_diff_by_op	signal_close	sma_10	hi_lo_diff_by_cl_l	sma_5	vol_op_l_ratio	vol_cl_l_ratio	vol_l_percent_diff	target	OOF_pred_glm	OOF_pred_nb	OOF_pred_gbm
577	788	2017-03-27	0.5941293	-0.0860625	0.0259496	-0.1089624	-0.1193624	0.0403788	0.0403788	0.4141655	0.4252867	-0.2157906	-0.7407335	0.0645776	0.6234565	-0.7380919	0.1003626	-0.7402292	0.0261690	0.0230661	0.0443486	No	0.1714786	0.0007146	0.0862415
578	807	2017-02-28	2.1115116	1.3386435	1.3890034	1.3505992	1.3935076	1.2993760	1.2993760	0.3519109	1.1673768	-0.3635380	-0.8372603	-1.0525550	1.3213788	-0.8182669	-1.1111413	-0.8128807	-0.5186544	-0.5190177	-0.5559561	No	0.0266172	0.0000018	0.0250923
579	963	2016-07-15	0.1820300	-0.1372888	-0.1295660	-0.0065412	-0.0005129	-0.1694323	-0.1694323	1.1565614	-0.2014732	-0.1091657	-1.5739526	-0.8234686	-0.4716169	-1.5579675	-0.9091749	-1.5415584	-0.2721179	-0.2725347	-0.2656148	No	0.0884190	0.0064820	0.3607964
580	341	2019-01-04	-1.6841555	-2.6985096	-2.8102498	-2.6479124	-2.9306403	-2.4113899	-2.4113899	-1.3951217	-3.1012486	-1.9059406	-0.4187683	1.2591131	-3.1829453	-0.5042487	0.8392387	-0.5215289	0.5518789	0.5316656	0.4342095	No	0.1401079	0.9999145	0.4710399
581	303	2019-03-01	0.4176277	0.2945936	0.1437523	0.3127654	0.3039570	0.3298467	0.3298467	0.6494030	0.5705581	-0.1129988	-0.1385437	-0.4132670	0.6092631	-0.1314229	-0.4058939	-0.1142257	-0.7394433	-0.7373897	-0.7674001	No	0.1186773	0.0285232	0.0621122
583	836	2017-01-17	0.9556489	-0.0365185	0.0596136	0.0017050	0.0000713	0.1339740	0.1339740	0.9078716	0.3346000	-0.6770169	-1.1610390	-0.0684980	0.3075284	-1.1550585	-0.0444153	-1.1376118	-0.5981109	-0.5929348	-0.6063222	No	0.0723700	0.0036941	0.0333305

evaluation (prediction) output

trend	symbol	n_period	l_period	date	target	pred_final_avg	pred_final
bearish	AAPL	5	20	2020-01-03	No	0.1861071	0
bearish	AAPL	5	20	2020-01-06	No	0.0357360	0
bearish	AAPL	5	20	2020-01-07	No	0.1161324	0
bearish	AAPL	5	20	2020-01-08	No	0.0312756	0
bearish	AAPL	5	20	2020-01-09	No	0.1627764	0
bearish	AAPL	5	20	2020-01-10	No	0.2898218	0
bearish	AAPL	5	20	2020-01-13	No	0.0300195	0
bearish	AAPL	5	20	2020-01-14	No	0.0406912	0
bearish	AAPL	5	20	2020-01-15	No	0.0304247	0
bearish	AAPL	5	20	2020-01-16	No	0.0386506	0
bearish	AAPL	5	20	2020-01-17	Yes	0.0877919	0
bearish	AAPL	5	20	2020-01-21	No	0.0425072	0
bearish	AAPL	5	20	2020-01-22	No	0.0851848	0
bearish	AAPL	5	20	2020-01-23	No	0.1274586	0
bearish	AAPL	5	20	2020-01-24	Yes	0.2583408	0
bearish	AAPL	5	20	2020-01-27	No	0.6950139	1
bearish	AAPL	5	20	2020-01-28	No	0.1140367	0
bearish	AAPL	5	20	2020-01-29	No	0.0322130	0
bearish	AAPL	5	20	2020-01-30	No	0.0338633	0
bearish	AAPL	5	20	2020-01-31	No	0.1093895	0
bearish	AAPL	5	20	2020-02-03	No	0.0337342	0
bearish	AAPL	5	20	2020-02-04	No	0.0308722	0
bearish	AAPL	5	20	2020-02-05	No	0.0369197	0
bearish	AAPL	5	20	2020-02-06	No	0.0333376	0
bearish	AAPL	5	20	2020-02-07	No	0.0297580	0
bearish	AAPL	5	20	2020-02-10	No	0.0531578	0
bearish	AAPL	5	20	2020-02-11	No	0.0277412	0
bearish	AAPL	5	20	2020-02-12	No	0.0296434	0
bearish	AAPL	5	20	2020-02-13	Yes	0.0323237	0
bearish	AAPL	5	20	2020-02-14	Yes	0.1231750	0
bearish	AAPL	5	20	2020-02-18	Yes	0.7426590	1
bearish	AAPL	5	20	2020-02-19	Yes	0.5701448	1
bearish	AAPL	5	20	2020-02-20	Yes	0.7746513	1
bearish	AAPL	5	20	2020-02-21	Yes	0.9082341	1
bearish	AAPL	5	20	2020-02-24	No	0.8194669	1
bearish	AAPL	5	20	2020-02-25	No	0.5853079	1
bearish	AAPL	5	20	2020-02-26	No	0.4502759	0
bearish	AAPL	5	20	2020-02-27	No	0.3688501	0
bearish	AAPL	5	20	2020-02-28	No	0.9561448	1
bearish	AAPL	5	20	2020-03-02	Yes	0.7279586	1
bearish	AAPL	5	20	2020-03-03	No	0.4082508	0
bearish	AAPL	5	20	2020-03-04	Yes	0.6489827	1
bearish	AAPL	5	20	2020-03-05	Yes	0.8656721	1
bearish	AAPL	5	20	2020-03-06	Yes	0.9364427	1
bearish	AAPL	5	20	2020-03-09	Yes	0.9453148	1
bearish	AAPL	5	20	2020-03-10	Yes	0.8233193	1
bearish	AAPL	5	20	2020-03-11	Yes	0.7021333	1
bearish	AAPL	5	20	2020-03-12	No	0.9597048	1
bearish	AAPL	5	20	2020-03-13	Yes	0.9364633	1
bearish	AAPL	5	20	2020-03-16	Yes	0.9384207	1
bearish	AAPL	5	20	2020-03-17	No	0.9211511	1
bearish	AAPL	5	20	2020-03-18	No	0.7667433	1
bearish	AAPL	5	20	2020-03-19	No	0.6779139	1
bearish	AAPL	5	20	2020-03-20	No	0.9292353	1
bearish	AAPL	5	20	2020-03-23	No	0.7835432	1
bearish	AAPL	5	20	2020-03-24	No	0.3590859	0
bearish	AAPL	5	20	2020-03-25	No	0.1895799	0
bearish	AAPL	5	20	2020-03-26	Yes	0.0841470	0
bearish	AAPL	5	20	2020-03-27	Yes	0.0287567	0
bearish	AAPL	5	20	2020-03-30	No	0.0694237	0
bearish	AAPL	5	20	2020-03-31	No	0.0341530	0
bearish	AAPL	5	20	2020-04-01	No	0.0341902	0
bearish	AAPL	5	20	2020-04-02	No	0.0737650	0
bearish	AAPL	5	20	2020-04-03	No	0.0343496	0
bearish	AAPL	5	20	2020-04-06	No	0.0289993	0
bearish	AAPL	5	20	2020-04-07	No	0.0287062	0
bearish	AAPL	5	20	2020-04-08	No	0.0270313	0
bearish	AAPL	5	20	2020-04-09	No	0.0744268	0
bearish	AAPL	5	20	2020-04-13	No	0.0320685	0
bearish	AAPL	5	20	2020-04-14	Yes	0.0274313	0
bearish	AAPL	5	20	2020-04-15	Yes	0.0276674	0
bearish	AAPL	5	20	2020-04-16	Yes	0.0269232	0
bearish	AAPL	5	20	2020-04-17	No	0.0274259	0
bearish	AAPL	5	20	2020-04-20	No	0.0276178	0
bearish	AAPL	5	20	2020-04-21	No	0.0276837	0
bearish	AAPL	5	20	2020-04-22	No	0.0288562	0
bearish	AAPL	5	20	2020-04-23	No	0.0340632	0
bearish	AAPL	5	20	2020-04-24	No	0.1854106	0
bearish	AAPL	5	20	2020-04-27	No	0.0299120	0
bearish	AAPL	5	20	2020-04-28	No	0.0325296	0
bearish	AAPL	5	20	2020-04-29	No	0.0267825	0
bearish	AAPL	5	20	2020-04-30	No	0.0284602	0
bearish	AAPL	5	20	2020-05-01	No	0.0333916	0
bearish	AAPL	5	20	2020-05-04	No	0.0508018	0
bearish	AAPL	5	20	2020-05-05	No	0.0414098	0
bearish	AAPL	5	20	2020-05-06	No	0.0794864	0
bearish	AAPL	5	20	2020-05-07	NA	0.0308495	0
bearish	AAPL	5	20	2020-05-08	NA	0.0283522	0
bearish	AAPL	5	20	2020-05-11	NA	0.0303203	0
bearish	AAPL	5	20	2020-05-12	NA	0.0273319	0
bearish	AAPL	5	20	2020-05-13	NA	0.0515707	0

model summary output

trend	symbol	n_period	l_period	trainMethod	accuracy	balanced_accuracy	detection_prevalence	detection_rate	f1	kappa	neg_pred_value	pos_pred_value	precision	prevalence	recall	sensitivity	specificity
bearish	AAPL	1	1	final - averaging	0.7889908	0.6970637	0.2018349	0.1376147	0.5660377	0.4319057	0.8160920	0.6818182	0.6818182	0.2844037	0.4838710	0.4838710	0.9102564
bearish	AAPL	1	5	final - averaging	0.7889908	0.7456576	0.2935780	0.1834862	0.6349206	0.4865861	0.8571429	0.6250000	0.6250000	0.2844037	0.6451613	0.6451613	0.8461538
bearish	AAPL	1	10	final - averaging	0.7870370	0.7397436	0.2870370	0.1759259	0.6229508	0.4746193	0.8571429	0.6129032	0.6129032	0.2777778	0.6333333	0.6333333	0.8461538
bearish	AAPL	1	15	final - averaging	0.8272727	0.7491987	0.2090909	0.1636364	0.6545455	0.5434688	0.8390805	0.7826087	0.7826087	0.2909091	0.5625000	0.5625000	0.9358974
bearish	AAPL	1	20	final - averaging	0.8490566	0.8516484	0.3396226	0.2264151	0.7500000	0.6442953	0.9428571	0.6666667	0.6666667	0.2641509	0.8571429	0.8571429	0.8461538
bearish	AAPL	1	1	gbm	0.7889908	0.6970637	0.2018349	0.1376147	0.5660377	0.4319057	0.8160920	0.6818182	0.6818182	0.2844037	0.4838710	0.4838710	0.9102564
bearish	AAPL	1	5	gbm	0.7981651	0.7520678	0.2844037	0.1834862	0.6451613	0.5041356	0.8589744	0.6451613	0.6451613	0.2844037	0.6451613	0.6451613	0.8589744
bearish	AAPL	1	10	gbm	0.7962963	0.7564103	0.2962963	0.1851852	0.6451613	0.5025126	0.8684211	0.6250000	0.6250000	0.2777778	0.6666667	0.6666667	0.8461538
bearish	AAPL	1	15	gbm	0.8272727	0.7584135	0.2272727	0.1727273	0.6666667	0.5524625	0.8470588	0.7600000	0.7600000	0.2909091	0.5937500	0.5937500	0.9230769
bearish	AAPL	1	20	gbm	0.8396226	0.8337912	0.3301887	0.2169811	0.7301587	0.6180585	0.9295775	0.6571429	0.6571429	0.2641509	0.8214286	0.8214286	0.8461538
bearish	AAPL	1	1	gbm - top layer	0.7798165	0.6809347	0.1926606	0.1284404	0.5384615	0.4008246	0.8068182	0.6666667	0.6666667	0.2844037	0.4516129	0.4516129	0.9102564
bearish	AAPL	1	5	gbm - top layer	0.8165138	0.7551696	0.2477064	0.1743119	0.6551724	0.5309811	0.8536585	0.7037037	0.7037037	0.2844037	0.6129032	0.6129032	0.8974359
bearish	AAPL	1	10	gbm - top layer	0.7777778	0.7230769	0.2777778	0.1666667	0.6000000	0.4461538	0.8461538	0.6000000	0.6000000	0.2777778	0.6000000	0.6000000	0.8461538
bearish	AAPL	1	15	gbm - top layer	0.7727273	0.6462340	0.1363636	0.1000000	0.4680851	0.3467933	0.7789474	0.7333333	0.7333333	0.2909091	0.3437500	0.3437500	0.9487179
bearish	AAPL	1	20	gbm - top layer	0.7735849	0.7316850	0.3018868	0.1698113	0.6000000	0.4430823	0.8648649	0.5625000	0.5625000	0.2641509	0.6428571	0.6428571	0.8205128
bearish	AAPL	1	1	glm	0.7981651	0.6743176	0.1376147	0.1100917	0.5217391	0.4128306	0.7978723	0.8000000	0.8000000	0.2844037	0.3870968	0.3870968	0.9615385
bearish	AAPL	1	5	glm	0.7522936	0.6519851	0.2018349	0.1192661	0.4905660	0.3331067	0.7931034	0.5909091	0.5909091	0.2844037	0.4193548	0.4193548	0.8846154
bearish	AAPL	1	10	glm	0.7685185	0.6756410	0.2129630	0.1296296	0.5283019	0.3784530	0.8117647	0.6086957	0.6086957	0.2777778	0.4666667	0.4666667	0.8846154
bearish	AAPL	1	15	glm	0.7454545	0.6177885	0.1454545	0.0909091	0.4166667	0.2763158	0.7659574	0.6250000	0.6250000	0.2909091	0.3125000	0.3125000	0.9230769
bearish	AAPL	1	20	glm	0.8113208	0.7573260	0.2641509	0.1698113	0.6428571	0.5146520	0.8717949	0.6428571	0.6428571	0.2641509	0.6428571	0.6428571	0.8717949
bearish	AAPL	1	1	glm - top layer	0.7889908	0.6873449	0.1834862	0.1284404	0.5490196	0.4195416	0.8089888	0.7000000	0.7000000	0.2844037	0.4516129	0.4516129	0.9230769
bearish	AAPL	1	5	glm - top layer	0.8073394	0.7487593	0.2568807	0.1743119	0.6440678	0.5124601	0.8518519	0.6785714	0.6785714	0.2844037	0.6129032	0.6129032	0.8846154
bearish	AAPL	1	10	glm - top layer	0.7962963	0.7564103	0.2962963	0.1851852	0.6451613	0.5025126	0.8684211	0.6250000	0.6250000	0.2777778	0.6666667	0.6666667	0.8461538
bearish	AAPL	1	15	glm - top layer	0.8090909	0.7271635	0.2090909	0.1545455	0.6181818	0.4954128	0.8275862	0.7391304	0.7391304	0.2909091	0.5312500	0.5312500	0.9230769
bearish	AAPL	1	20	glm - top layer	0.8490566	0.8402015	0.3207547	0.2169811	0.7419355	0.6366752	0.9305556	0.6764706	0.6764706	0.2641509	0.8214286	0.8214286	0.8589744
bearish	AAPL	1	1	nb	0.7614679	0.6875517	0.2477064	0.1467890	0.5517241	0.3902754	0.8170732	0.5925926	0.5925926	0.2844037	0.5161290	0.5161290	0.8589744
bearish	AAPL	1	5	nb	0.6880734	0.6459884	0.3394495	0.1559633	0.5000000	0.2758890	0.8055556	0.4594595	0.4594595	0.2844037	0.5483871	0.5483871	0.7435897
bearish	AAPL	1	10	nb	0.7222222	0.6743590	0.3148148	0.1574074	0.5312500	0.3349754	0.8243243	0.5000000	0.5000000	0.2777778	0.5666667	0.5666667	0.7820513
bearish	AAPL	1	15	nb	0.7000000	0.6225962	0.2636364	0.1272727	0.4590164	0.2521632	0.7777778	0.4827586	0.4827586	0.2909091	0.4375000	0.4375000	0.8076923
bearish	AAPL	1	20	nb	0.7830189	0.7495421	0.3113208	0.1792453	0.6229508	0.4720658	0.8767123	0.5757576	0.5757576	0.2641509	0.6785714	0.6785714	0.8205128
bearish	AAPL	1	1	nb - top layer	0.7798165	0.7003722	0.2293578	0.1467890	0.5714286	0.4255599	0.8214286	0.6400000	0.6400000	0.2844037	0.5161290	0.5161290	0.8846154
bearish	AAPL	1	5	nb - top layer	0.7614679	0.7264268	0.3211009	0.1834862	0.6060606	0.4359076	0.8513514	0.5714286	0.5714286	0.2844037	0.6451613	0.6451613	0.8076923
bearish	AAPL	1	10	nb - top layer	0.7500000	0.7346154	0.3611111	0.1944444	0.6086957	0.4295775	0.8695652	0.5384615	0.5384615	0.2777778	0.7000000	0.7000000	0.7692308
bearish	AAPL	1	15	nb - top layer	0.8000000	0.7207532	0.2181818	0.1545455	0.6071429	0.4766436	0.8255814	0.7083333	0.7083333	0.2909091	0.5312500	0.5312500	0.9102564
bearish	AAPL	1	20	nb - top layer	0.8396226	0.8223443	0.3113208	0.2075472	0.7213115	0.6097878	0.9178082	0.6666667	0.6666667	0.2641509	0.7857143	0.7857143	0.8589744
bearish	AAPL	2	1	final - averaging	0.7818182	0.7355769	0.2909091	0.1818182	0.6250000	0.4711538	0.8461538	0.6250000	0.6250000	0.2909091	0.6250000	0.6250000	0.8461538
bearish	AAPL	2	5	final - averaging	0.8288288	0.7995338	0.3063063	0.2162162	0.7164179	0.5938764	0.8831169	0.7058824	0.7058824	0.2972973	0.7272727	0.7272727	0.8717949
bearish	AAPL	2	10	final - averaging	0.8035714	0.7760181	0.3214286	0.2142857	0.6857143	0.5430267	0.8684211	0.6666667	0.6666667	0.3035714	0.7058824	0.7058824	0.8461538
bearish	AAPL	2	15	final - averaging	0.8648649	0.8251748	0.2702703	0.2162162	0.7619048	0.6678636	0.8888889	0.8000000	0.8000000	0.2972973	0.7272727	0.7272727	0.9230769
bearish	AAPL	2	20	final - averaging	0.8571429	0.8310709	0.3035714	0.2321429	0.7647059	0.6621418	0.8974359	0.7647059	0.7647059	0.3035714	0.7647059	0.7647059	0.8974359
bearish	AAPL	2	1	gbm	0.8363636	0.7924679	0.2727273	0.2000000	0.7096774	0.5959184	0.8750000	0.7333333	0.7333333	0.2909091	0.6875000	0.6875000	0.8974359
bearish	AAPL	2	5	gbm	0.8468468	0.8210956	0.3063063	0.2252252	0.7462687	0.6366262	0.8961039	0.7352941	0.7352941	0.2972973	0.7575758	0.7575758	0.8846154
bearish	AAPL	2	10	gbm	0.7857143	0.7466063	0.3035714	0.1964286	0.6470588	0.4932127	0.8461538	0.6470588	0.6470588	0.3035714	0.6470588	0.6470588	0.8461538
bearish	AAPL	2	15	gbm	0.8648649	0.8251748	0.2702703	0.2162162	0.7619048	0.6678636	0.8888889	0.8000000	0.8000000	0.2972973	0.7272727	0.7272727	0.9230769
bearish	AAPL	2	20	gbm	0.8660714	0.8291855	0.2767857	0.2232143	0.7692308	0.6751740	0.8888889	0.8064516	0.8064516	0.3035714	0.7352941	0.7352941	0.9230769
bearish	AAPL	2	1	gbm - top layer	0.8636364	0.8209135	0.2636364	0.2090909	0.7540984	0.6600742	0.8888889	0.7931034	0.7931034	0.2909091	0.7187500	0.7187500	0.9230769
bearish	AAPL	2	5	gbm - top layer	0.8198198	0.8018648	0.3333333	0.2252252	0.7142857	0.5833333	0.8918919	0.6756757	0.6756757	0.2972973	0.7575758	0.7575758	0.8461538
bearish	AAPL	2	10	gbm - top layer	0.8035714	0.7594268	0.2857143	0.1964286	0.6666667	0.5276074	0.8500000	0.6875000	0.6875000	0.3035714	0.6470588	0.6470588	0.8717949
bearish	AAPL	2	15	gbm - top layer	0.8738739	0.8403263	0.2792793	0.2252252	0.7812500	0.6927639	0.9000000	0.8064516	0.8064516	0.2972973	0.7575758	0.7575758	0.9230769
bearish	AAPL	2	20	gbm - top layer	0.8571429	0.8310709	0.3035714	0.2321429	0.7647059	0.6621418	0.8974359	0.7647059	0.7647059	0.3035714	0.7647059	0.7647059	0.8974359
bearish	AAPL	2	1	glm	0.6818182	0.5637019	0.1909091	0.0818182	0.3396226	0.1417744	0.7415730	0.4285714	0.4285714	0.2909091	0.2812500	0.2812500	0.8461538
bearish	AAPL	2	5	glm	0.7747748	0.6736597	0.1801802	0.1261261	0.5283019	0.3918475	0.7912088	0.7000000	0.7000000	0.2972973	0.4242424	0.4242424	0.9230769
bearish	AAPL	2	10	glm	0.7589286	0.7024887	0.2767857	0.1696429	0.5846154	0.4153132	0.8148148	0.6129032	0.6129032	0.3035714	0.5588235	0.5588235	0.8461538
bearish	AAPL	2	15	glm	0.7567568	0.6608392	0.1981982	0.1261261	0.5090909	0.3558994	0.7865169	0.6363636	0.6363636	0.2972973	0.4242424	0.4242424	0.8974359
bearish	AAPL	2	20	glm	0.7142857	0.6206637	0.2142857	0.1160714	0.4482759	0.2631579	0.7613636	0.5416667	0.5416667	0.3035714	0.3823529	0.3823529	0.8589744
bearish	AAPL	2	1	glm - top layer	0.8181818	0.7520032	0.2363636	0.1727273	0.6551724	0.5335030	0.8452381	0.7307692	0.7307692	0.2909091	0.5937500	0.5937500	0.9102564
bearish	AAPL	2	5	glm - top layer	0.8288288	0.7907925	0.2882883	0.2072072	0.7076923	0.5867137	0.8734177	0.7187500	0.7187500	0.2972973	0.6969697	0.6969697	0.8846154
bearish	AAPL	2	10	glm - top layer	0.8035714	0.7594268	0.2857143	0.1964286	0.6666667	0.5276074	0.8500000	0.6875000	0.6875000	0.3035714	0.6470588	0.6470588	0.8717949
bearish	AAPL	2	15	glm - top layer	0.8648649	0.8251748	0.2702703	0.2162162	0.7619048	0.6678636	0.8888889	0.8000000	0.8000000	0.2972973	0.7272727	0.7272727	0.9230769
bearish	AAPL	2	20	glm - top layer	0.8482143	0.7997738	0.2589286	0.2053571	0.7301587	0.6254917	0.8674699	0.7931034	0.7931034	0.3035714	0.6764706	0.6764706	0.9230769
bearish	AAPL	2	1	nb	0.6363636	0.5500801	0.2727273	0.1000000	0.3548387	0.1020408	0.7375000	0.3666667	0.3666667	0.2909091	0.3437500	0.3437500	0.7564103
bearish	AAPL	2	5	nb	0.7207207	0.6789044	0.3243243	0.1711712	0.5507246	0.3486655	0.8133333	0.5277778	0.5277778	0.2972973	0.5757576	0.5757576	0.7820513
bearish	AAPL	2	10	nb	0.7053571	0.6474359	0.2946429	0.1517857	0.5074627	0.2973384	0.7848101	0.5151515	0.5151515	0.3035714	0.5000000	0.5000000	0.7948718
bearish	AAPL	2	15	nb	0.6486486	0.5926573	0.3243243	0.1351351	0.4347826	0.1805792	0.7600000	0.4166667	0.4166667	0.2972973	0.4545455	0.4545455	0.7307692
bearish	AAPL	2	20	nb	0.6875000	0.6843891	0.4196429	0.2053571	0.5679012	0.3328795	0.8307692	0.4893617	0.4893617	0.3035714	0.6764706	0.6764706	0.6923077
bearish	AAPL	2	1	nb - top layer	0.7636364	0.7227564	0.3090909	0.1818182	0.6060606	0.4374508	0.8421053	0.5882353	0.5882353	0.2909091	0.6250000	0.6250000	0.8205128
bearish	AAPL	2	5	nb - top layer	0.7747748	0.7610723	0.3603604	0.2162162	0.6575342	0.4920373	0.8732394	0.6000000	0.6000000	0.2972973	0.7272727	0.7272727	0.7948718
bearish	AAPL	2	10	nb - top layer	0.7946429	0.7696078	0.3303571	0.2142857	0.6760563	0.5261221	0.8666667	0.6486486	0.6486486	0.3035714	0.7058824	0.7058824	0.8333333
bearish	AAPL	2	15	nb - top layer	0.8648649	0.8339161	0.2882883	0.2252252	0.7692308	0.6737213	0.8987342	0.7812500	0.7812500	0.2972973	0.7575758	0.7575758	0.9102564
bearish	AAPL	2	20	nb - top layer	0.8214286	0.8054299	0.3392857	0.2321429	0.7222222	0.5912409	0.8918919	0.6842105	0.6842105	0.3035714	0.7647059	0.7647059	0.8461538
bearish	AAPL	3	1	final - averaging	0.8425926	0.8192308	0.3055556	0.2129630	0.7301587	0.6194030	0.9066667	0.6969697	0.6969697	0.2777778	0.7666667	0.7666667	0.8717949
bearish	AAPL	3	5	final - averaging	0.8508772	0.8162393	0.2894737	0.2280702	0.7536232	0.6469945	0.8765432	0.7878788	0.7878788	0.3157895	0.7222222	0.7222222	0.9102564
bearish	AAPL	3	10	final - averaging	0.8198198	0.7843823	0.2972973	0.2072072	0.6969697	0.5687646	0.8717949	0.6969697	0.6969697	0.2972973	0.6969697	0.6969697	0.8717949
bearish	AAPL	3	15	final - averaging	0.8440367	0.8035567	0.2752294	0.2018349	0.7213115	0.6130716	0.8860759	0.7333333	0.7333333	0.2844037	0.7096774	0.7096774	0.8974359
bearish	AAPL	3	20	final - averaging	0.8807339	0.8583540	0.2935780	0.2293578	0.7936508	0.7098095	0.9220779	0.7812500	0.7812500	0.2844037	0.8064516	0.8064516	0.9102564
bearish	AAPL	3	1	gbm	0.8703704	0.8487179	0.2962963	0.2222222	0.7741935	0.6834171	0.9210526	0.7500000	0.7500000	0.2777778	0.8000000	0.8000000	0.8974359
bearish	AAPL	3	5	gbm	0.8508772	0.8087607	0.2719298	0.2192982	0.7462687	0.6415094	0.8674699	0.8064516	0.8064516	0.3157895	0.6944444	0.6944444	0.9230769
bearish	AAPL	3	10	gbm	0.8198198	0.7756410	0.2792793	0.1981982	0.6875000	0.5610913	0.8625000	0.7096774	0.7096774	0.2972973	0.6666667	0.6666667	0.8846154
bearish	AAPL	3	15	gbm	0.8532110	0.8099669	0.2660550	0.2018349	0.7333333	0.6322227	0.8875000	0.7586207	0.7586207	0.2844037	0.7096774	0.7096774	0.9102564
bearish	AAPL	3	20	gbm	0.8899083	0.8744830	0.3027523	0.2385321	0.8125000	0.7346856	0.9342105	0.7878788	0.7878788	0.2844037	0.8387097	0.8387097	0.9102564
bearish	AAPL	3	1	gbm - top layer	0.8240741	0.7961538	0.3055556	0.2037037	0.6984127	0.5746269	0.8933333	0.6666667	0.6666667	0.2777778	0.7333333	0.7333333	0.8589744
bearish	AAPL	3	5	gbm - top layer	0.8421053	0.8023504	0.2807018	0.2192982	0.7352941	0.6233480	0.8658537	0.7812500	0.7812500	0.3157895	0.6944444	0.6944444	0.9102564
bearish	AAPL	3	10	gbm - top layer	0.8198198	0.7843823	0.2972973	0.2072072	0.6969697	0.5687646	0.8717949	0.6969697	0.6969697	0.2972973	0.6969697	0.6969697	0.8717949
bearish	AAPL	3	15	gbm - top layer	0.8532110	0.8196857	0.2844037	0.2110092	0.7419355	0.6393714	0.8974359	0.7419355	0.7419355	0.2844037	0.7419355	0.7419355	0.8974359
bearish	AAPL	3	20	gbm - top layer	0.8807339	0.8583540	0.2935780	0.2293578	0.7936508	0.7098095	0.9220779	0.7812500	0.7812500	0.2844037	0.8064516	0.8064516	0.9102564
bearish	AAPL	3	1	glm	0.7314815	0.5884615	0.1388889	0.0740741	0.3555556	0.2090909	0.7634409	0.5333333	0.5333333	0.2777778	0.2666667	0.2666667	0.9102564
bearish	AAPL	3	5	glm	0.7280702	0.6292735	0.1842105	0.1140351	0.4561404	0.2912154	0.7526882	0.6190476	0.6190476	0.3157895	0.3611111	0.3611111	0.8974359
bearish	AAPL	3	10	glm	0.7207207	0.6089744	0.1801802	0.0990991	0.4150943	0.2458909	0.7582418	0.5500000	0.5500000	0.2972973	0.3333333	0.3333333	0.8846154
bearish	AAPL	3	15	glm	0.7706422	0.6842432	0.2201835	0.1376147	0.5454545	0.3953850	0.8117647	0.6250000	0.6250000	0.2844037	0.4838710	0.4838710	0.8846154
bearish	AAPL	3	20	glm	0.7706422	0.6842432	0.2201835	0.1376147	0.5454545	0.3953850	0.8117647	0.6250000	0.6250000	0.2844037	0.4838710	0.4838710	0.8846154
bearish	AAPL	3	1	glm - top layer	0.8518519	0.8256410	0.2962963	0.2129630	0.7419355	0.6381910	0.9078947	0.7187500	0.7187500	0.2777778	0.7666667	0.7666667	0.8846154
bearish	AAPL	3	5	glm - top layer	0.8684211	0.8290598	0.2719298	0.2280702	0.7761194	0.6836848	0.8795181	0.8387097	0.8387097	0.3157895	0.7222222	0.7222222	0.9358974
bearish	AAPL	3	10	glm - top layer	0.8198198	0.7756410	0.2792793	0.1981982	0.6875000	0.5610913	0.8625000	0.7096774	0.7096774	0.2972973	0.6666667	0.6666667	0.8846154
bearish	AAPL	3	15	glm - top layer	0.8532110	0.8099669	0.2660550	0.2018349	0.7333333	0.6322227	0.8875000	0.7586207	0.7586207	0.2844037	0.7096774	0.7096774	0.9102564
bearish	AAPL	3	20	glm - top layer	0.8807339	0.8583540	0.2935780	0.2293578	0.7936508	0.7098095	0.9220779	0.7812500	0.7812500	0.2844037	0.8064516	0.8064516	0.9102564
bearish	AAPL	3	1	nb	0.7037037	0.6615385	0.3333333	0.1574074	0.5151515	0.3043478	0.8194444	0.4722222	0.4722222	0.2777778	0.5666667	0.5666667	0.7564103
bearish	AAPL	3	5	nb	0.6929825	0.6036325	0.2192982	0.1140351	0.4262295	0.2258440	0.7415730	0.5200000	0.5200000	0.3157895	0.3611111	0.3611111	0.8461538
bearish	AAPL	3	10	nb	0.6846847	0.6270396	0.3063063	0.1441441	0.4776119	0.2518775	0.7792208	0.4705882	0.4705882	0.2972973	0.4848485	0.4848485	0.7692308
bearish	AAPL	3	15	nb	0.7247706	0.6716294	0.3027523	0.1559633	0.5312500	0.3367140	0.8157895	0.5151515	0.5151515	0.2844037	0.5483871	0.5483871	0.7948718
bearish	AAPL	3	20	nb	0.6788991	0.6492969	0.3669725	0.1651376	0.5070423	0.2745769	0.8115942	0.4500000	0.4500000	0.2844037	0.5806452	0.5806452	0.7179487
bearish	AAPL	3	1	nb - top layer	0.8240741	0.8064103	0.3240741	0.2129630	0.7076923	0.5829268	0.9041096	0.6571429	0.6571429	0.2777778	0.7666667	0.7666667	0.8461538
bearish	AAPL	3	5	nb - top layer	0.7894737	0.7339744	0.2631579	0.1842105	0.6363636	0.4899329	0.8214286	0.7000000	0.7000000	0.3157895	0.5833333	0.5833333	0.8846154
bearish	AAPL	3	10	nb - top layer	0.7747748	0.7348485	0.3063063	0.1891892	0.6268657	0.4656268	0.8441558	0.6176471	0.6176471	0.2972973	0.6363636	0.6363636	0.8333333
bearish	AAPL	3	15	nb - top layer	0.8256881	0.7907361	0.2935780	0.2018349	0.6984127	0.5758755	0.8831169	0.6875000	0.6875000	0.2844037	0.7096774	0.7096774	0.8717949
bearish	AAPL	3	20	nb - top layer	0.8256881	0.7907361	0.2935780	0.2018349	0.6984127	0.5758755	0.8831169	0.6875000	0.6875000	0.2844037	0.7096774	0.7096774	0.8717949
bearish	AAPL	4	1	final - averaging	0.8495575	0.8043956	0.2654867	0.2123894	0.7384615	0.6337464	0.8674699	0.8000000	0.8000000	0.3097345	0.6857143	0.6857143	0.9230769
bearish	AAPL	4	5	final - averaging	0.8545455	0.8145032	0.2727273	0.2090909	0.7419355	0.6408163	0.8875000	0.7666667	0.7666667	0.2909091	0.7187500	0.7187500	0.9102564
bearish	AAPL	4	10	final - averaging	0.8333333	0.7820513	0.2592593	0.1851852	0.6896552	0.5759162	0.8750000	0.7142857	0.7142857	0.2777778	0.6666667	0.6666667	0.8974359
bearish	AAPL	4	15	final - averaging	0.8285714	0.7877493	0.2761905	0.1809524	0.6785714	0.5618915	0.8947368	0.6551724	0.6551724	0.2571429	0.7037037	0.7037037	0.8717949
bearish	AAPL	4	20	final - averaging	0.8518519	0.8051282	0.2592593	0.1944444	0.7241379	0.6230366	0.8875000	0.7500000	0.7500000	0.2777778	0.7000000	0.7000000	0.9102564
bearish	AAPL	4	1	gbm	0.8495575	0.7965201	0.2477876	0.2035398	0.7301587	0.6276410	0.8588235	0.8214286	0.8214286	0.3097345	0.6571429	0.6571429	0.9358974
bearish	AAPL	4	5	gbm	0.8545455	0.8145032	0.2727273	0.2090909	0.7419355	0.6408163	0.8875000	0.7666667	0.7666667	0.2909091	0.7187500	0.7187500	0.9102564
bearish	AAPL	4	10	gbm	0.8703704	0.8282051	0.2592593	0.2037037	0.7586207	0.6701571	0.9000000	0.7857143	0.7857143	0.2777778	0.7333333	0.7333333	0.9230769
bearish	AAPL	4	15	gbm	0.8380952	0.7941595	0.2666667	0.1809524	0.6909091	0.5812808	0.8961039	0.6785714	0.6785714	0.2571429	0.7037037	0.7037037	0.8846154
bearish	AAPL	4	20	gbm	0.8611111	0.8115385	0.2500000	0.1944444	0.7368421	0.6428571	0.8888889	0.7777778	0.7777778	0.2777778	0.7000000	0.7000000	0.9230769
bearish	AAPL	4	1	gbm - top layer	0.8584071	0.8108059	0.2566372	0.2123894	0.7500000	0.6524414	0.8690476	0.8275862	0.8275862	0.3097345	0.6857143	0.6857143	0.9358974
bearish	AAPL	4	5	gbm - top layer	0.8454545	0.7988782	0.2636364	0.2000000	0.7213115	0.6147507	0.8765432	0.7586207	0.7586207	0.2909091	0.6875000	0.6875000	0.9102564
bearish	AAPL	4	10	gbm - top layer	0.8796296	0.8346154	0.2500000	0.2037037	0.7719298	0.6904762	0.9012346	0.8148148	0.8148148	0.2777778	0.7333333	0.7333333	0.9358974
bearish	AAPL	4	15	gbm - top layer	0.8285714	0.7514245	0.2190476	0.1523810	0.6400000	0.5284431	0.8658537	0.6956522	0.6956522	0.2571429	0.5925926	0.5925926	0.9102564
bearish	AAPL	4	20	gbm - top layer	0.8796296	0.8141026	0.2129630	0.1851852	0.7547170	0.6767956	0.8823529	0.8695652	0.8695652	0.2777778	0.6666667	0.6666667	0.9615385
bearish	AAPL	4	1	glm	0.7079646	0.5915751	0.1592920	0.0884956	0.3773585	0.2114612	0.7368421	0.5555556	0.5555556	0.3097345	0.2857143	0.2857143	0.8974359
bearish	AAPL	4	5	glm	0.7727273	0.6738782	0.1909091	0.1272727	0.5283019	0.3869817	0.7977528	0.6666667	0.6666667	0.2909091	0.4375000	0.4375000	0.9102564
bearish	AAPL	4	10	glm	0.7314815	0.5987179	0.1574074	0.0833333	0.3829787	0.2278107	0.7692308	0.5294118	0.5294118	0.2777778	0.3000000	0.3000000	0.8974359
bearish	AAPL	4	15	glm	0.7809524	0.6467236	0.1523810	0.0952381	0.4651163	0.3385374	0.8089888	0.6250000	0.6250000	0.2571429	0.3703704	0.3703704	0.9230769
bearish	AAPL	4	20	glm	0.7500000	0.6628205	0.2314815	0.1296296	0.5090909	0.3432432	0.8072289	0.5600000	0.5600000	0.2777778	0.4666667	0.4666667	0.8589744
bearish	AAPL	4	1	glm - top layer	0.8495575	0.8043956	0.2654867	0.2123894	0.7384615	0.6337464	0.8674699	0.8000000	0.8000000	0.3097345	0.6857143	0.6857143	0.9230769
bearish	AAPL	4	5	glm - top layer	0.8454545	0.7988782	0.2636364	0.2000000	0.7213115	0.6147507	0.8765432	0.7586207	0.7586207	0.2909091	0.6875000	0.6875000	0.9102564
bearish	AAPL	4	10	glm - top layer	0.8703704	0.8282051	0.2592593	0.2037037	0.7586207	0.6701571	0.9000000	0.7857143	0.7857143	0.2777778	0.7333333	0.7333333	0.9230769
bearish	AAPL	4	15	glm - top layer	0.8380952	0.7820513	0.2476190	0.1714286	0.6792453	0.5710166	0.8860759	0.6923077	0.6923077	0.2571429	0.6666667	0.6666667	0.8974359
bearish	AAPL	4	20	glm - top layer	0.8425926	0.7782051	0.2314815	0.1759259	0.6909091	0.5864865	0.8674699	0.7600000	0.7600000	0.2777778	0.6333333	0.6333333	0.9230769
bearish	AAPL	4	1	nb	0.7079646	0.6152015	0.2123894	0.1150442	0.4406780	0.2522559	0.7528090	0.5416667	0.5416667	0.3097345	0.3714286	0.3714286	0.8589744
bearish	AAPL	4	5	nb	0.7272727	0.6786859	0.3090909	0.1636364	0.5454545	0.3509048	0.8157895	0.5294118	0.5294118	0.2909091	0.5625000	0.5625000	0.7948718
bearish	AAPL	4	10	nb	0.7314815	0.6089744	0.1759259	0.0925926	0.4081633	0.2456647	0.7752809	0.5263158	0.5263158	0.2777778	0.3333333	0.3333333	0.8846154
bearish	AAPL	4	15	nb	0.7142857	0.6381766	0.2761905	0.1238095	0.4642857	0.2698192	0.8157895	0.4482759	0.4482759	0.2571429	0.4814815	0.4814815	0.7948718
bearish	AAPL	4	20	nb	0.7962963	0.7564103	0.2962963	0.1851852	0.6451613	0.5025126	0.8684211	0.6250000	0.6250000	0.2777778	0.6666667	0.6666667	0.8461538
bearish	AAPL	4	1	nb - top layer	0.8141593	0.7708791	0.2831858	0.2035398	0.6865672	0.5548678	0.8518519	0.7187500	0.7187500	0.3097345	0.6571429	0.6571429	0.8846154
bearish	AAPL	4	5	nb - top layer	0.8454545	0.8173077	0.3000000	0.2181818	0.7384615	0.6288210	0.8961039	0.7272727	0.7272727	0.2909091	0.7500000	0.7500000	0.8846154
bearish	AAPL	4	10	nb - top layer	0.8148148	0.7589744	0.2592593	0.1759259	0.6551724	0.5287958	0.8625000	0.6785714	0.6785714	0.2777778	0.6333333	0.6333333	0.8846154
bearish	AAPL	4	15	nb - top layer	0.8190476	0.8055556	0.3238095	0.2000000	0.6885246	0.5633618	0.9154930	0.6176471	0.6176471	0.2571429	0.7777778	0.7777778	0.8333333
bearish	AAPL	4	20	nb - top layer	0.8148148	0.7794872	0.2962963	0.1944444	0.6774194	0.5477387	0.8815789	0.6562500	0.6562500	0.2777778	0.7000000	0.7000000	0.8589744
bearish	AAPL	5	1	final - averaging	0.8035714	0.7760181	0.3214286	0.2142857	0.6857143	0.5430267	0.8684211	0.6666667	0.6666667	0.3035714	0.7058824	0.7058824	0.8461538
bearish	AAPL	5	5	final - averaging	0.8141593	0.7787546	0.3008850	0.2123894	0.6956522	0.5619347	0.8607595	0.7058824	0.7058824	0.3097345	0.6857143	0.6857143	0.8717949
bearish	AAPL	5	10	final - averaging	0.8584071	0.8108059	0.2566372	0.2123894	0.7500000	0.6524414	0.8690476	0.8275862	0.8275862	0.3097345	0.6857143	0.6857143	0.9358974
bearish	AAPL	5	15	final - averaging	0.8611111	0.8217949	0.2685185	0.2037037	0.7457627	0.6502591	0.8987342	0.7586207	0.7586207	0.2777778	0.7333333	0.7333333	0.9102564
bearish	AAPL	5	20	final - averaging	0.8363636	0.8293269	0.3454545	0.2363636	0.7428571	0.6241458	0.9166667	0.6842105	0.6842105	0.2909091	0.8125000	0.8125000	0.8461538
bearish	AAPL	5	1	gbm	0.8035714	0.7594268	0.2857143	0.1964286	0.6666667	0.5276074	0.8500000	0.6875000	0.6875000	0.3035714	0.6470588	0.6470588	0.8717949
bearish	AAPL	5	5	gbm	0.7964602	0.7501832	0.2831858	0.1946903	0.6567164	0.5124742	0.8395062	0.6875000	0.6875000	0.3097345	0.6285714	0.6285714	0.8717949
bearish	AAPL	5	10	gbm	0.8672566	0.8172161	0.2477876	0.2123894	0.7619048	0.6714480	0.8705882	0.8571429	0.8571429	0.3097345	0.6857143	0.6857143	0.9487179
bearish	AAPL	5	15	gbm	0.8796296	0.8346154	0.2500000	0.2037037	0.7719298	0.6904762	0.9012346	0.8148148	0.8148148	0.2777778	0.7333333	0.7333333	0.9358974
bearish	AAPL	5	20	gbm	0.8272727	0.8229167	0.3545455	0.2363636	0.7323944	0.6066993	0.9154930	0.6666667	0.6666667	0.2909091	0.8125000	0.8125000	0.8333333
bearish	AAPL	5	1	gbm - top layer	0.7946429	0.7613122	0.3125000	0.2053571	0.6666667	0.5183246	0.8571429	0.6571429	0.6571429	0.3035714	0.6764706	0.6764706	0.8461538
bearish	AAPL	5	5	gbm - top layer	0.8141593	0.7708791	0.2831858	0.2035398	0.6865672	0.5548678	0.8518519	0.7187500	0.7187500	0.3097345	0.6571429	0.6571429	0.8846154
bearish	AAPL	5	10	gbm - top layer	0.8584071	0.7950549	0.2212389	0.1946903	0.7333333	0.6405567	0.8522727	0.8800000	0.8800000	0.3097345	0.6285714	0.6285714	0.9615385
bearish	AAPL	5	15	gbm - top layer	0.8611111	0.8217949	0.2685185	0.2037037	0.7457627	0.6502591	0.8987342	0.7586207	0.7586207	0.2777778	0.7333333	0.7333333	0.9102564
bearish	AAPL	5	20	gbm - top layer	0.8181818	0.8072917	0.3454545	0.2272727	0.7142857	0.5823842	0.9027778	0.6578947	0.6578947	0.2909091	0.7812500	0.7812500	0.8333333
bearish	AAPL	5	1	glm	0.7232143	0.6602564	0.2767857	0.1517857	0.5230769	0.3286930	0.7901235	0.5483871	0.5483871	0.3035714	0.5000000	0.5000000	0.8205128
bearish	AAPL	5	5	glm	0.7256637	0.6358974	0.2123894	0.1238938	0.4745763	0.2975737	0.7640449	0.5833333	0.5833333	0.3097345	0.4000000	0.4000000	0.8717949
bearish	AAPL	5	10	glm	0.7345133	0.6265568	0.1681416	0.1061947	0.4444444	0.2896060	0.7553191	0.6315789	0.6315789	0.3097345	0.3428571	0.3428571	0.9102564
bearish	AAPL	5	15	glm	0.8055556	0.7012821	0.1759259	0.1296296	0.5714286	0.4537572	0.8202247	0.7368421	0.7368421	0.2777778	0.4666667	0.4666667	0.9358974
bearish	AAPL	5	20	glm	0.7636364	0.6766827	0.2181818	0.1363636	0.5357143	0.3814879	0.8023256	0.6250000	0.6250000	0.2909091	0.4687500	0.4687500	0.8846154
bearish	AAPL	5	1	glm - top layer	0.8125000	0.7658371	0.2767857	0.1964286	0.6769231	0.5452436	0.8518519	0.7096774	0.7096774	0.3035714	0.6470588	0.6470588	0.8846154
bearish	AAPL	5	5	glm - top layer	0.8053097	0.7644689	0.2920354	0.2035398	0.6764706	0.5374023	0.8500000	0.6969697	0.6969697	0.3097345	0.6571429	0.6571429	0.8717949
bearish	AAPL	5	10	glm - top layer	0.8672566	0.8172161	0.2477876	0.2123894	0.7619048	0.6714480	0.8705882	0.8571429	0.8571429	0.3097345	0.6857143	0.6857143	0.9487179
bearish	AAPL	5	15	glm - top layer	0.8796296	0.8243590	0.2314815	0.1944444	0.7636364	0.6837838	0.8915663	0.8400000	0.8400000	0.2777778	0.7000000	0.7000000	0.9487179
bearish	AAPL	5	20	glm - top layer	0.8363636	0.8293269	0.3454545	0.2363636	0.7428571	0.6241458	0.9166667	0.6842105	0.6842105	0.2909091	0.8125000	0.8125000	0.8461538
bearish	AAPL	5	1	nb	0.7053571	0.6723228	0.3482143	0.1785714	0.5479452	0.3309196	0.8082192	0.5128205	0.5128205	0.3035714	0.5882353	0.5882353	0.7564103
bearish	AAPL	5	5	nb	0.7168142	0.6609890	0.2920354	0.1592920	0.5294118	0.3271306	0.7875000	0.5454545	0.5454545	0.3097345	0.5142857	0.5142857	0.8076923
bearish	AAPL	5	10	nb	0.6991150	0.6324176	0.2743363	0.1415929	0.4848485	0.2734493	0.7682927	0.5161290	0.5161290	0.3097345	0.4571429	0.4571429	0.8076923
bearish	AAPL	5	15	nb	0.7962963	0.7461538	0.2777778	0.1759259	0.6333333	0.4923077	0.8589744	0.6333333	0.6333333	0.2777778	0.6333333	0.6333333	0.8589744
bearish	AAPL	5	20	nb	0.7181818	0.6538462	0.2818182	0.1454545	0.5079365	0.3105540	0.7974684	0.5161290	0.5161290	0.2909091	0.5000000	0.5000000	0.8076923
bearish	AAPL	5	1	nb - top layer	0.7857143	0.7631976	0.3392857	0.2142857	0.6666667	0.5094891	0.8648649	0.6315789	0.6315789	0.3035714	0.7058824	0.7058824	0.8205128
bearish	AAPL	5	5	nb - top layer	0.7610619	0.7324176	0.3362832	0.2035398	0.6301370	0.4541063	0.8400000	0.6052632	0.6052632	0.3097345	0.6571429	0.6571429	0.8076923
bearish	AAPL	5	10	nb - top layer	0.8141593	0.7551282	0.2477876	0.1858407	0.6666667	0.5400271	0.8352941	0.7500000	0.7500000	0.3097345	0.6000000	0.6000000	0.9102564
bearish	AAPL	5	15	nb - top layer	0.8518519	0.8051282	0.2592593	0.1944444	0.7241379	0.6230366	0.8875000	0.7500000	0.7500000	0.2777778	0.7000000	0.7000000	0.9102564
bearish	AAPL	5	20	nb - top layer	0.8272727	0.8137019	0.3363636	0.2272727	0.7246377	0.5997702	0.9041096	0.6756757	0.6756757	0.2909091	0.7812500	0.7812500	0.8461538
bullish	AAPL	1	1	final - averaging	0.8240741	0.7653846	0.2500000	0.1759259	0.6666667	0.5476190	0.8641975	0.7037037	0.7037037	0.2777778	0.6333333	0.6333333	0.8974359
bullish	AAPL	1	5	final - averaging	0.8518519	0.7743590	0.2037037	0.1666667	0.6923077	0.5977654	0.8604651	0.8181818	0.8181818	0.2777778	0.6000000	0.6000000	0.9487179
bullish	AAPL	1	10	final - averaging	0.8018018	0.7628205	0.2972973	0.1981982	0.6666667	0.5256410	0.8589744	0.6666667	0.6666667	0.2972973	0.6666667	0.6666667	0.8589744
bullish	AAPL	1	15	final - averaging	0.8198198	0.8106061	0.3513514	0.2342342	0.7222222	0.5902547	0.9027778	0.6666667	0.6666667	0.2972973	0.7878788	0.7878788	0.8333333
bullish	AAPL	1	20	final - averaging	0.8545455	0.8052885	0.2545455	0.2000000	0.7333333	0.6339434	0.8780488	0.7857143	0.7857143	0.2909091	0.6875000	0.6875000	0.9230769
bullish	AAPL	1	1	gbm	0.8240741	0.7653846	0.2500000	0.1759259	0.6666667	0.5476190	0.8641975	0.7037037	0.7037037	0.2777778	0.6333333	0.6333333	0.8974359
bullish	AAPL	1	5	gbm	0.8611111	0.7910256	0.2129630	0.1759259	0.7169811	0.6270718	0.8705882	0.8260870	0.8260870	0.2777778	0.6333333	0.6333333	0.9487179
bullish	AAPL	1	10	gbm	0.8108108	0.7692308	0.2882883	0.1981982	0.6769231	0.5432099	0.8607595	0.6875000	0.6875000	0.2972973	0.6666667	0.6666667	0.8717949
bullish	AAPL	1	15	gbm	0.8378378	0.8234266	0.3333333	0.2342342	0.7428571	0.6250000	0.9054054	0.7027027	0.7027027	0.2972973	0.7878788	0.7878788	0.8589744
bullish	AAPL	1	20	gbm	0.8545455	0.8145032	0.2727273	0.2090909	0.7419355	0.6408163	0.8875000	0.7666667	0.7666667	0.2909091	0.7187500	0.7187500	0.9102564
bullish	AAPL	1	1	gbm - top layer	0.8333333	0.7615385	0.2222222	0.1666667	0.6666667	0.5573770	0.8571429	0.7500000	0.7500000	0.2777778	0.6000000	0.6000000	0.9230769
bullish	AAPL	1	5	gbm - top layer	0.8333333	0.7410256	0.1851852	0.1481481	0.6400000	0.5371429	0.8409091	0.8000000	0.8000000	0.2777778	0.5333333	0.5333333	0.9487179
bullish	AAPL	1	10	gbm - top layer	0.8198198	0.7668998	0.2612613	0.1891892	0.6774194	0.5531401	0.8536585	0.7241379	0.7241379	0.2972973	0.6363636	0.6363636	0.8974359
bullish	AAPL	1	15	gbm - top layer	0.8198198	0.7843823	0.2972973	0.2072072	0.6969697	0.5687646	0.8717949	0.6969697	0.6969697	0.2972973	0.6969697	0.6969697	0.8717949
bullish	AAPL	1	20	gbm - top layer	0.8818182	0.8337340	0.2454545	0.2090909	0.7796610	0.6997060	0.8915663	0.8518519	0.8518519	0.2909091	0.7187500	0.7187500	0.9487179
bullish	AAPL	1	1	glm	0.7685185	0.6346154	0.1388889	0.0925926	0.4444444	0.3181818	0.7849462	0.6666667	0.6666667	0.2777778	0.3333333	0.3333333	0.9358974
bullish	AAPL	1	5	glm	0.7314815	0.5987179	0.1574074	0.0833333	0.3829787	0.2278107	0.7692308	0.5294118	0.5294118	0.2777778	0.3000000	0.3000000	0.8974359
bullish	AAPL	1	10	glm	0.7297297	0.6416084	0.2252252	0.1261261	0.4827586	0.3045113	0.7790698	0.5600000	0.5600000	0.2972973	0.4242424	0.4242424	0.8589744
bullish	AAPL	1	15	glm	0.7027027	0.6136364	0.2342342	0.1171171	0.4406780	0.2420857	0.7647059	0.5000000	0.5000000	0.2972973	0.3939394	0.3939394	0.8333333
bullish	AAPL	1	20	glm	0.7545455	0.6149840	0.1181818	0.0818182	0.4000000	0.2787761	0.7628866	0.6923077	0.6923077	0.2909091	0.2812500	0.2812500	0.9487179
bullish	AAPL	1	1	glm - top layer	0.8333333	0.7717949	0.2407407	0.1759259	0.6785714	0.5668449	0.8658537	0.7307692	0.7307692	0.2777778	0.6333333	0.6333333	0.9102564
bullish	AAPL	1	5	glm - top layer	0.8518519	0.7743590	0.2037037	0.1666667	0.6923077	0.5977654	0.8604651	0.8181818	0.8181818	0.2777778	0.6000000	0.6000000	0.9487179
bullish	AAPL	1	10	glm - top layer	0.8108108	0.7692308	0.2882883	0.1981982	0.6769231	0.5432099	0.8607595	0.6875000	0.6875000	0.2972973	0.6666667	0.6666667	0.8717949
bullish	AAPL	1	15	glm - top layer	0.8378378	0.8146853	0.3153153	0.2252252	0.7352941	0.6185567	0.8947368	0.7142857	0.7142857	0.2972973	0.7575758	0.7575758	0.8717949
bullish	AAPL	1	20	glm - top layer	0.8545455	0.8052885	0.2545455	0.2000000	0.7333333	0.6339434	0.8780488	0.7857143	0.7857143	0.2909091	0.6875000	0.6875000	0.9230769
bullish	AAPL	1	1	nb	0.7222222	0.6641026	0.2962963	0.1481481	0.5161290	0.3216080	0.8157895	0.5000000	0.5000000	0.2777778	0.5333333	0.5333333	0.7948718
bullish	AAPL	1	5	nb	0.7314815	0.6910256	0.3240741	0.1666667	0.5538462	0.3634146	0.8356164	0.5142857	0.5142857	0.2777778	0.6000000	0.6000000	0.7820513
bullish	AAPL	1	10	nb	0.6936937	0.6509324	0.3333333	0.1621622	0.5142857	0.2916667	0.7972973	0.4864865	0.4864865	0.2972973	0.5454545	0.5454545	0.7564103
bullish	AAPL	1	15	nb	0.6936937	0.6684149	0.3693694	0.1801802	0.5405405	0.3148148	0.8142857	0.4878049	0.4878049	0.2972973	0.6060606	0.6060606	0.7307692
bullish	AAPL	1	20	nb	0.6909091	0.6622596	0.3636364	0.1727273	0.5277778	0.3022388	0.8142857	0.4750000	0.4750000	0.2909091	0.5937500	0.5937500	0.7307692
bullish	AAPL	1	1	nb - top layer	0.8055556	0.7628205	0.2870370	0.1851852	0.6557377	0.5203046	0.8701299	0.6451613	0.6451613	0.2777778	0.6666667	0.6666667	0.8589744
bullish	AAPL	1	5	nb - top layer	0.8055556	0.7423077	0.2500000	0.1666667	0.6315789	0.5000000	0.8518519	0.6666667	0.6666667	0.2777778	0.6000000	0.6000000	0.8846154
bullish	AAPL	1	10	nb - top layer	0.7297297	0.6940559	0.3333333	0.1801802	0.5714286	0.3750000	0.8243243	0.5405405	0.5405405	0.2972973	0.6060606	0.6060606	0.7820513
bullish	AAPL	1	15	nb - top layer	0.7747748	0.7610723	0.3603604	0.2162162	0.6575342	0.4920373	0.8732394	0.6000000	0.6000000	0.2972973	0.7272727	0.7272727	0.7948718
bullish	AAPL	1	20	nb - top layer	0.8454545	0.8080929	0.2818182	0.2090909	0.7301587	0.6219167	0.8860759	0.7419355	0.7419355	0.2909091	0.7187500	0.7187500	0.8974359
bullish	AAPL	2	1	final - averaging	0.8148148	0.7487179	0.2407407	0.1666667	0.6428571	0.5187166	0.8536585	0.6923077	0.6923077	0.2777778	0.6000000	0.6000000	0.8974359
bullish	AAPL	2	5	final - averaging	0.7927928	0.7476690	0.2882883	0.1891892	0.6461538	0.4997061	0.8481013	0.6562500	0.6562500	0.2972973	0.6363636	0.6363636	0.8589744
bullish	AAPL	2	10	final - averaging	0.8348624	0.7582713	0.2110092	0.1651376	0.6666667	0.5600897	0.8488372	0.7826087	0.7826087	0.2844037	0.5806452	0.5806452	0.9358974
bullish	AAPL	2	15	final - averaging	0.8090909	0.7455929	0.2454545	0.1727273	0.6440678	0.5149097	0.8433735	0.7037037	0.7037037	0.2909091	0.5937500	0.5937500	0.8974359
bullish	AAPL	2	20	final - averaging	0.8148148	0.7589744	0.2592593	0.1759259	0.6551724	0.5287958	0.8625000	0.6785714	0.6785714	0.2777778	0.6333333	0.6333333	0.8846154
bullish	AAPL	2	1	gbm	0.7870370	0.7294872	0.2685185	0.1666667	0.6101695	0.4637306	0.8481013	0.6206897	0.6206897	0.2777778	0.6000000	0.6000000	0.8589744
bullish	AAPL	2	5	gbm	0.7927928	0.7389277	0.2702703	0.1801802	0.6349206	0.4907241	0.8395062	0.6666667	0.6666667	0.2972973	0.6060606	0.6060606	0.8717949
bullish	AAPL	2	10	gbm	0.8073394	0.7390405	0.2385321	0.1651376	0.6315789	0.5024995	0.8433735	0.6923077	0.6923077	0.2844037	0.5806452	0.5806452	0.8974359
bullish	AAPL	2	15	gbm	0.8000000	0.7299679	0.2363636	0.1636364	0.6206897	0.4868533	0.8333333	0.6923077	0.6923077	0.2909091	0.5625000	0.5625000	0.8974359
bullish	AAPL	2	20	gbm	0.8148148	0.7692308	0.2777778	0.1851852	0.6666667	0.5384615	0.8717949	0.6666667	0.6666667	0.2777778	0.6666667	0.6666667	0.8717949
bullish	AAPL	2	1	gbm - top layer	0.7870370	0.6987179	0.2129630	0.1388889	0.5660377	0.4281768	0.8235294	0.6521739	0.6521739	0.2777778	0.5000000	0.5000000	0.8974359
bullish	AAPL	2	5	gbm - top layer	0.8198198	0.7668998	0.2612613	0.1891892	0.6774194	0.5531401	0.8536585	0.7241379	0.7241379	0.2972973	0.6363636	0.6363636	0.8974359
bullish	AAPL	2	10	gbm - top layer	0.8256881	0.7421423	0.2018349	0.1559633	0.6415094	0.5307047	0.8390805	0.7727273	0.7727273	0.2844037	0.5483871	0.5483871	0.9358974
bullish	AAPL	2	15	gbm - top layer	0.8000000	0.7299679	0.2363636	0.1636364	0.6206897	0.4868533	0.8333333	0.6923077	0.6923077	0.2909091	0.5625000	0.5625000	0.8974359
bullish	AAPL	2	20	gbm - top layer	0.8240741	0.7756410	0.2685185	0.1851852	0.6779661	0.5569948	0.8734177	0.6896552	0.6896552	0.2777778	0.6666667	0.6666667	0.8846154
bullish	AAPL	2	1	glm	0.7314815	0.5884615	0.1388889	0.0740741	0.3555556	0.2090909	0.7634409	0.5333333	0.5333333	0.2777778	0.2666667	0.2666667	0.9102564
bullish	AAPL	2	5	glm	0.7117117	0.6113054	0.2072072	0.1081081	0.4285714	0.2439336	0.7613636	0.5217391	0.5217391	0.2972973	0.3636364	0.3636364	0.8589744
bullish	AAPL	2	10	glm	0.7339450	0.5905707	0.1284404	0.0733945	0.3555556	0.2169928	0.7578947	0.5714286	0.5714286	0.2844037	0.2580645	0.2580645	0.9230769
bullish	AAPL	2	15	glm	0.7181818	0.5801282	0.1363636	0.0727273	0.3404255	0.1900238	0.7473684	0.5333333	0.5333333	0.2909091	0.2500000	0.2500000	0.9102564
bullish	AAPL	2	20	glm	0.7222222	0.6025641	0.1851852	0.0925926	0.4000000	0.2285714	0.7727273	0.5000000	0.5000000	0.2777778	0.3333333	0.3333333	0.8717949
bullish	AAPL	2	1	glm - top layer	0.8055556	0.7423077	0.2500000	0.1666667	0.6315789	0.5000000	0.8518519	0.6666667	0.6666667	0.2777778	0.6000000	0.6000000	0.8846154
bullish	AAPL	2	5	glm - top layer	0.7927928	0.7301865	0.2522523	0.1711712	0.6229508	0.4814138	0.8313253	0.6785714	0.6785714	0.2972973	0.5757576	0.5757576	0.8846154
bullish	AAPL	2	10	glm - top layer	0.8256881	0.7421423	0.2018349	0.1559633	0.6415094	0.5307047	0.8390805	0.7727273	0.7727273	0.2844037	0.5483871	0.5483871	0.9358974
bullish	AAPL	2	15	glm - top layer	0.8090909	0.7455929	0.2454545	0.1727273	0.6440678	0.5149097	0.8433735	0.7037037	0.7037037	0.2909091	0.5937500	0.5937500	0.8974359
bullish	AAPL	2	20	glm - top layer	0.8240741	0.7756410	0.2685185	0.1851852	0.6779661	0.5569948	0.8734177	0.6896552	0.6896552	0.2777778	0.6666667	0.6666667	0.8846154
bullish	AAPL	2	1	nb	0.7037037	0.6205128	0.2592593	0.1203704	0.4482759	0.2460733	0.7875000	0.4642857	0.4642857	0.2777778	0.4333333	0.4333333	0.8076923
bullish	AAPL	2	5	nb	0.6486486	0.6276224	0.3963964	0.1711712	0.4935065	0.2328549	0.7910448	0.4318182	0.4318182	0.2972973	0.5757576	0.5757576	0.6794872
bullish	AAPL	2	10	nb	0.6513761	0.5814723	0.3027523	0.1192661	0.4062500	0.1598377	0.7631579	0.3939394	0.3939394	0.2844037	0.4193548	0.4193548	0.7435897
bullish	AAPL	2	15	nb	0.6727273	0.5665064	0.2181818	0.0909091	0.3571429	0.1435986	0.7441860	0.4166667	0.4166667	0.2909091	0.3125000	0.3125000	0.8205128
bullish	AAPL	2	20	nb	0.6944444	0.6858974	0.3981481	0.1851852	0.5479452	0.3280543	0.8461538	0.4651163	0.4651163	0.2777778	0.6666667	0.6666667	0.7051282
bullish	AAPL	2	1	nb - top layer	0.7962963	0.7564103	0.2962963	0.1851852	0.6451613	0.5025126	0.8684211	0.6250000	0.6250000	0.2777778	0.6666667	0.6666667	0.8461538
bullish	AAPL	2	5	nb - top layer	0.7387387	0.7092075	0.3423423	0.1891892	0.5915493	0.4008934	0.8356164	0.5526316	0.5526316	0.2972973	0.6363636	0.6363636	0.7820513
bullish	AAPL	2	10	nb - top layer	0.7981651	0.7423490	0.2660550	0.1743119	0.6333333	0.4943062	0.8500000	0.6551724	0.6551724	0.2844037	0.6129032	0.6129032	0.8717949
bullish	AAPL	2	15	nb - top layer	0.7636364	0.6951122	0.2545455	0.1545455	0.5666667	0.4051581	0.8170732	0.6071429	0.6071429	0.2909091	0.5312500	0.5312500	0.8589744
bullish	AAPL	2	20	nb - top layer	0.8333333	0.7923077	0.2777778	0.1944444	0.7000000	0.5846154	0.8846154	0.7000000	0.7000000	0.2777778	0.7000000	0.7000000	0.8846154
bullish	AAPL	3	1	final - averaging	0.8018018	0.7453380	0.2612613	0.1801802	0.6451613	0.5084541	0.8414634	0.6896552	0.6896552	0.2972973	0.6060606	0.6060606	0.8846154
bullish	AAPL	3	5	final - averaging	0.7657658	0.6672494	0.1891892	0.1261261	0.5185185	0.3736979	0.7888889	0.6666667	0.6666667	0.2972973	0.4242424	0.4242424	0.9102564
bullish	AAPL	3	10	final - averaging	0.7272727	0.6418269	0.2363636	0.1272727	0.4827586	0.3002545	0.7857143	0.5384615	0.5384615	0.2909091	0.4375000	0.4375000	0.8461538
bullish	AAPL	3	15	final - averaging	0.7747748	0.6736597	0.1801802	0.1261261	0.5283019	0.3918475	0.7912088	0.7000000	0.7000000	0.2972973	0.4242424	0.4242424	0.9230769
bullish	AAPL	3	20	final - averaging	0.7719298	0.7136752	0.2631579	0.1754386	0.6060606	0.4474273	0.8095238	0.6666667	0.6666667	0.3157895	0.5555556	0.5555556	0.8717949
bullish	AAPL	3	1	gbm	0.8108108	0.7604895	0.2702703	0.1891892	0.6666667	0.5350090	0.8518519	0.7000000	0.7000000	0.2972973	0.6363636	0.6363636	0.8846154
bullish	AAPL	3	5	gbm	0.7387387	0.6829837	0.2882883	0.1621622	0.5538462	0.3691946	0.8101266	0.5625000	0.5625000	0.2972973	0.5454545	0.5454545	0.8205128
bullish	AAPL	3	10	gbm	0.7545455	0.6702724	0.2272727	0.1363636	0.5263158	0.3640257	0.8000000	0.6000000	0.6000000	0.2909091	0.4687500	0.4687500	0.8717949
bullish	AAPL	3	15	gbm	0.7927928	0.7039627	0.1981982	0.1441441	0.5818182	0.4513217	0.8089888	0.7272727	0.7272727	0.2972973	0.4848485	0.4848485	0.9230769
bullish	AAPL	3	20	gbm	0.7807018	0.7200855	0.2543860	0.1754386	0.6153846	0.4644870	0.8117647	0.6896552	0.6896552	0.3157895	0.5555556	0.5555556	0.8846154
bullish	AAPL	3	1	gbm - top layer	0.8378378	0.7709790	0.2252252	0.1801802	0.6896552	0.5827068	0.8488372	0.8000000	0.8000000	0.2972973	0.6060606	0.6060606	0.9358974
bullish	AAPL	3	5	gbm - top layer	0.7837838	0.6713287	0.1531532	0.1171171	0.5200000	0.3983740	0.7872340	0.7647059	0.7647059	0.2972973	0.3939394	0.3939394	0.9487179
bullish	AAPL	3	10	gbm - top layer	0.7090909	0.6197917	0.2363636	0.1181818	0.4482759	0.2536047	0.7738095	0.5000000	0.5000000	0.2909091	0.4062500	0.4062500	0.8333333
bullish	AAPL	3	15	gbm - top layer	0.8108108	0.6993007	0.1441441	0.1261261	0.5714286	0.4681725	0.8000000	0.8750000	0.8750000	0.2972973	0.4242424	0.4242424	0.9743590
bullish	AAPL	3	20	gbm - top layer	0.7894737	0.7264957	0.2456140	0.1754386	0.6250000	0.4818182	0.8139535	0.7142857	0.7142857	0.3157895	0.5555556	0.5555556	0.8974359
bullish	AAPL	3	1	glm	0.7297297	0.5804196	0.0990991	0.0630631	0.3181818	0.1991342	0.7400000	0.6363636	0.6363636	0.2972973	0.2121212	0.2121212	0.9487179
bullish	AAPL	3	5	glm	0.7027027	0.5699301	0.1441441	0.0720721	0.3265306	0.1642710	0.7368421	0.5000000	0.5000000	0.2972973	0.2424242	0.2424242	0.8974359
bullish	AAPL	3	10	glm	0.6454545	0.5012019	0.1545455	0.0454545	0.2040816	0.0027894	0.7096774	0.2941176	0.2941176	0.2909091	0.1562500	0.1562500	0.8461538
bullish	AAPL	3	15	glm	0.6756757	0.5769231	0.2252252	0.0990991	0.3793103	0.1654135	0.7441860	0.4400000	0.4400000	0.2972973	0.3333333	0.3333333	0.8205128
bullish	AAPL	3	20	glm	0.7192982	0.5929487	0.1228070	0.0789474	0.3600000	0.2225064	0.7300000	0.6428571	0.6428571	0.3157895	0.2500000	0.2500000	0.9358974
bullish	AAPL	3	1	glm - top layer	0.8288288	0.7733100	0.2522523	0.1891892	0.6885246	0.5716027	0.8554217	0.7500000	0.7500000	0.2972973	0.6363636	0.6363636	0.9102564
bullish	AAPL	3	5	glm - top layer	0.7387387	0.6655012	0.2522523	0.1441441	0.5245902	0.3461304	0.7951807	0.5714286	0.5714286	0.2972973	0.4848485	0.4848485	0.8461538
bullish	AAPL	3	10	glm - top layer	0.7545455	0.6702724	0.2272727	0.1363636	0.5263158	0.3640257	0.8000000	0.6000000	0.6000000	0.2909091	0.4687500	0.4687500	0.8717949
bullish	AAPL	3	15	glm - top layer	0.8108108	0.7167832	0.1801802	0.1441441	0.6037736	0.4891519	0.8131868	0.8000000	0.8000000	0.2972973	0.4848485	0.4848485	0.9487179
bullish	AAPL	3	20	glm - top layer	0.7982456	0.7403846	0.2543860	0.1842105	0.6461538	0.5073281	0.8235294	0.7241379	0.7241379	0.3157895	0.5833333	0.5833333	0.8974359
bullish	AAPL	3	1	nb	0.7027027	0.5961538	0.1981982	0.0990991	0.4000000	0.2127660	0.7528090	0.5000000	0.5000000	0.2972973	0.3333333	0.3333333	0.8589744
bullish	AAPL	3	5	nb	0.7027027	0.6223776	0.2522523	0.1261261	0.4590164	0.2559415	0.7710843	0.5000000	0.5000000	0.2972973	0.4242424	0.4242424	0.8205128
bullish	AAPL	3	10	nb	0.6454545	0.5933494	0.3363636	0.1363636	0.4347826	0.1784757	0.7671233	0.4054054	0.4054054	0.2909091	0.4687500	0.4687500	0.7179487
bullish	AAPL	3	15	nb	0.6666667	0.5879953	0.2702703	0.1171171	0.4126984	0.1807301	0.7530864	0.4333333	0.4333333	0.2972973	0.3939394	0.3939394	0.7820513
bullish	AAPL	3	20	nb	0.6403509	0.5876068	0.3245614	0.1403509	0.4383562	0.1739130	0.7402597	0.4324324	0.4324324	0.3157895	0.4444444	0.4444444	0.7307692
bullish	AAPL	3	1	nb - top layer	0.7657658	0.7022145	0.2612613	0.1621622	0.5806452	0.4190821	0.8170732	0.6206897	0.6206897	0.2972973	0.5454545	0.5454545	0.8589744
bullish	AAPL	3	5	nb - top layer	0.7387387	0.6742424	0.2702703	0.1531532	0.5396825	0.3578695	0.8024691	0.5666667	0.5666667	0.2972973	0.5151515	0.5151515	0.8333333
bullish	AAPL	3	10	nb - top layer	0.6909091	0.6069712	0.2545455	0.1181818	0.4333333	0.2221298	0.7682927	0.4642857	0.4642857	0.2909091	0.4062500	0.4062500	0.8076923
bullish	AAPL	3	15	nb - top layer	0.7477477	0.6544289	0.2072072	0.1261261	0.5000000	0.3384419	0.7840909	0.6086957	0.6086957	0.2972973	0.4242424	0.4242424	0.8846154
bullish	AAPL	3	20	nb - top layer	0.7631579	0.7147436	0.2894737	0.1842105	0.6086957	0.4393443	0.8148148	0.6363636	0.6363636	0.3157895	0.5833333	0.5833333	0.8461538
bullish	AAPL	4	1	final - averaging	0.8378378	0.8117089	0.3063063	0.2162162	0.7272727	0.6120388	0.8961039	0.7058824	0.7058824	0.2882883	0.7500000	0.7500000	0.8734177
bullish	AAPL	4	5	final - averaging	0.8055556	0.7579660	0.2777778	0.1759259	0.6440678	0.5103627	0.8717949	0.6333333	0.6333333	0.2685185	0.6551724	0.6551724	0.8607595
bullish	AAPL	4	10	final - averaging	0.8301887	0.8373183	0.3490566	0.2169811	0.7187500	0.6013372	0.9420290	0.6216216	0.6216216	0.2547170	0.8518519	0.8518519	0.8227848
bullish	AAPL	4	15	final - averaging	0.8198198	0.7432753	0.2162162	0.1621622	0.6428571	0.5256410	0.8390805	0.7500000	0.7500000	0.2882883	0.5625000	0.5625000	0.9240506
bullish	AAPL	4	20	final - averaging	0.7962963	0.7298123	0.2500000	0.1574074	0.6071429	0.4698795	0.8518519	0.6296296	0.6296296	0.2685185	0.5862069	0.5862069	0.8734177
bullish	AAPL	4	1	gbm	0.8378378	0.8024130	0.2882883	0.2072072	0.7187500	0.6048259	0.8860759	0.7187500	0.7187500	0.2882883	0.7187500	0.7187500	0.8860759
bullish	AAPL	4	5	gbm	0.8148148	0.7642951	0.2685185	0.1759259	0.6551724	0.5285901	0.8734177	0.6551724	0.6551724	0.2685185	0.6551724	0.6551724	0.8734177
bullish	AAPL	4	10	gbm	0.8490566	0.8377872	0.3113208	0.2075472	0.7333333	0.6295325	0.9315068	0.6666667	0.6666667	0.2547170	0.8148148	0.8148148	0.8607595
bullish	AAPL	4	15	gbm	0.8198198	0.7525712	0.2342342	0.1711712	0.6551724	0.5349811	0.8470588	0.7307692	0.7307692	0.2882883	0.5937500	0.5937500	0.9113924
bullish	AAPL	4	20	gbm	0.8148148	0.7533828	0.2500000	0.1666667	0.6428571	0.5180723	0.8641975	0.6666667	0.6666667	0.2685185	0.6206897	0.6206897	0.8860759
bullish	AAPL	4	1	gbm - top layer	0.8198198	0.7897547	0.3063063	0.2072072	0.6969697	0.5689320	0.8831169	0.6764706	0.6764706	0.2882883	0.7187500	0.7187500	0.8607595
bullish	AAPL	4	5	gbm - top layer	0.8055556	0.7579660	0.2777778	0.1759259	0.6440678	0.5103627	0.8717949	0.6333333	0.6333333	0.2685185	0.6551724	0.6551724	0.8607595
bullish	AAPL	4	10	gbm - top layer	0.7924528	0.7754337	0.3301887	0.1886792	0.6451613	0.5019223	0.9014085	0.5714286	0.5714286	0.2547170	0.7407407	0.7407407	0.8101266
bullish	AAPL	4	15	gbm - top layer	0.8018018	0.7213212	0.2162162	0.1531532	0.6071429	0.4782051	0.8275862	0.7083333	0.7083333	0.2882883	0.5312500	0.5312500	0.9113924
bullish	AAPL	4	20	gbm - top layer	0.8611111	0.8068529	0.2407407	0.1851852	0.7272727	0.6344765	0.8902439	0.7692308	0.7692308	0.2685185	0.6896552	0.6896552	0.9240506
bullish	AAPL	4	1	glm	0.7027027	0.5122627	0.0450450	0.0180180	0.1081081	0.0327436	0.7169811	0.4000000	0.4000000	0.2882883	0.0625000	0.0625000	0.9620253
bullish	AAPL	4	5	glm	0.7500000	0.5781318	0.0925926	0.0555556	0.3076923	0.1971366	0.7653061	0.6000000	0.6000000	0.2685185	0.2068966	0.2068966	0.9493671
bullish	AAPL	4	10	glm	0.6886792	0.4864041	0.0943396	0.0188679	0.1081081	-0.0342992	0.7395833	0.2000000	0.2000000	0.2547170	0.0740741	0.0740741	0.8987342
bullish	AAPL	4	15	glm	0.6756757	0.5025712	0.0900901	0.0270270	0.1428571	0.0064644	0.7128713	0.3000000	0.3000000	0.2882883	0.0937500	0.0937500	0.9113924
bullish	AAPL	4	20	glm	0.6666667	0.5102575	0.1574074	0.0462963	0.2173913	0.0236062	0.7362637	0.2941176	0.2941176	0.2685185	0.1724138	0.1724138	0.8481013
bullish	AAPL	4	1	glm - top layer	0.8288288	0.7867880	0.2792793	0.1981982	0.6984127	0.5789579	0.8750000	0.7096774	0.7096774	0.2882883	0.6875000	0.6875000	0.8860759
bullish	AAPL	4	5	glm - top layer	0.7962963	0.7407246	0.2685185	0.1666667	0.6206897	0.4814491	0.8607595	0.6206897	0.6206897	0.2685185	0.6206897	0.6206897	0.8607595
bullish	AAPL	4	10	glm - top layer	0.8679245	0.8626348	0.3113208	0.2169811	0.7666667	0.6758410	0.9452055	0.6969697	0.6969697	0.2547170	0.8518519	0.8518519	0.8734177
bullish	AAPL	4	15	glm - top layer	0.8108108	0.7369462	0.2252252	0.1621622	0.6315789	0.5068754	0.8372093	0.7200000	0.7200000	0.2882883	0.5625000	0.5625000	0.9113924
bullish	AAPL	4	20	glm - top layer	0.7962963	0.7189000	0.2314815	0.1481481	0.5925926	0.4577818	0.8433735	0.6400000	0.6400000	0.2685185	0.5517241	0.5517241	0.8860759
bullish	AAPL	4	1	nb	0.6936937	0.6174842	0.2702703	0.1261261	0.4516129	0.2394196	0.7777778	0.4666667	0.4666667	0.2882883	0.4375000	0.4375000	0.7974684
bullish	AAPL	4	5	nb	0.7407407	0.6263640	0.1944444	0.1018519	0.4400000	0.2769010	0.7931034	0.5238095	0.5238095	0.2685185	0.3793103	0.3793103	0.8734177
bullish	AAPL	4	10	nb	0.6320755	0.6190811	0.4150943	0.1509434	0.4507042	0.1972816	0.8225806	0.3636364	0.3636364	0.2547170	0.5925926	0.5925926	0.6455696
bullish	AAPL	4	15	nb	0.7027027	0.6331092	0.2792793	0.1351351	0.4761905	0.2687163	0.7875000	0.4838710	0.4838710	0.2882883	0.4687500	0.4687500	0.7974684
bullish	AAPL	4	20	nb	0.6296296	0.5940637	0.3796296	0.1388889	0.4285714	0.1663450	0.7910448	0.3658537	0.3658537	0.2685185	0.5172414	0.5172414	0.6708861
bullish	AAPL	4	1	nb - top layer	0.8018018	0.7585047	0.2882883	0.1891892	0.6562500	0.5170095	0.8607595	0.6562500	0.6562500	0.2882883	0.6562500	0.6562500	0.8607595
bullish	AAPL	4	5	nb - top layer	0.7962963	0.7516368	0.2870370	0.1759259	0.6333333	0.4925246	0.8701299	0.6129032	0.6129032	0.2685185	0.6551724	0.6551724	0.8481013
bullish	AAPL	4	10	nb - top layer	0.7924528	0.7876231	0.3490566	0.1981132	0.6562500	0.5127455	0.9130435	0.5675676	0.5675676	0.2547170	0.7777778	0.7777778	0.7974684
bullish	AAPL	4	15	nb - top layer	0.7657658	0.7053006	0.2702703	0.1621622	0.5806452	0.4183797	0.8271605	0.6000000	0.6000000	0.2882883	0.5625000	0.5625000	0.8481013
bullish	AAPL	4	20	nb - top layer	0.7314815	0.6855085	0.3148148	0.1574074	0.5396825	0.3518212	0.8378378	0.5000000	0.5000000	0.2685185	0.5862069	0.5862069	0.7848101
bullish	AAPL	5	1	final - averaging	0.7894737	0.7115385	0.2105263	0.1578947	0.6000000	0.4647887	0.8000000	0.7500000	0.7500000	0.3157895	0.5000000	0.5000000	0.9230769
bullish	AAPL	5	5	final - averaging	0.8157895	0.7756410	0.2894737	0.2105263	0.6956522	0.5639344	0.8518519	0.7272727	0.7272727	0.3157895	0.6666667	0.6666667	0.8846154
bullish	AAPL	5	10	final - averaging	0.8611111	0.7910256	0.2129630	0.1759259	0.7169811	0.6270718	0.8705882	0.8260870	0.8260870	0.2777778	0.6333333	0.6333333	0.9487179
bullish	AAPL	5	15	final - averaging	0.7410714	0.6564857	0.2232143	0.1339286	0.5084746	0.3382233	0.7816092	0.6000000	0.6000000	0.3035714	0.4411765	0.4411765	0.8717949
bullish	AAPL	5	20	final - averaging	0.8518519	0.7948718	0.2407407	0.1851852	0.7142857	0.6149733	0.8780488	0.7692308	0.7692308	0.2777778	0.6666667	0.6666667	0.9230769
bullish	AAPL	5	1	gbm	0.7807018	0.7126068	0.2368421	0.1666667	0.6031746	0.4558992	0.8045977	0.7037037	0.7037037	0.3157895	0.5277778	0.5277778	0.8974359
bullish	AAPL	5	5	gbm	0.8333333	0.7884615	0.2719298	0.2105263	0.7164179	0.5993341	0.8554217	0.7741935	0.7741935	0.3157895	0.6666667	0.6666667	0.9102564
bullish	AAPL	5	10	gbm	0.8703704	0.8076923	0.2222222	0.1851852	0.7407407	0.6557377	0.8809524	0.8333333	0.8333333	0.2777778	0.6666667	0.6666667	0.9487179
bullish	AAPL	5	15	gbm	0.7678571	0.6840121	0.2142857	0.1428571	0.5517241	0.4013158	0.7954545	0.6666667	0.6666667	0.3035714	0.4705882	0.4705882	0.8974359
bullish	AAPL	5	20	gbm	0.8518519	0.7948718	0.2407407	0.1851852	0.7142857	0.6149733	0.8780488	0.7692308	0.7692308	0.2777778	0.6666667	0.6666667	0.9230769
bullish	AAPL	5	1	gbm - top layer	0.8245614	0.7521368	0.2105263	0.1754386	0.6666667	0.5539906	0.8222222	0.8333333	0.8333333	0.3157895	0.5555556	0.5555556	0.9487179
bullish	AAPL	5	5	gbm - top layer	0.8421053	0.8173077	0.3157895	0.2368421	0.7500000	0.6346154	0.8846154	0.7500000	0.7500000	0.3157895	0.7500000	0.7500000	0.8846154
bullish	AAPL	5	10	gbm - top layer	0.8425926	0.7474359	0.1759259	0.1481481	0.6530612	0.5578035	0.8426966	0.8421053	0.8421053	0.2777778	0.5333333	0.5333333	0.9615385
bullish	AAPL	5	15	gbm - top layer	0.7500000	0.6711916	0.2321429	0.1428571	0.5333333	0.3667205	0.7906977	0.6153846	0.6153846	0.3035714	0.4705882	0.4705882	0.8717949
bullish	AAPL	5	20	gbm - top layer	0.8611111	0.8115385	0.2500000	0.1944444	0.7368421	0.6428571	0.8888889	0.7777778	0.7777778	0.2777778	0.7000000	0.7000000	0.9230769
bullish	AAPL	5	1	glm	0.6842105	0.5149573	0.0350877	0.0175439	0.1000000	0.0393258	0.6909091	0.5000000	0.5000000	0.3157895	0.0555556	0.0555556	0.9743590
bullish	AAPL	5	5	glm	0.6929825	0.5438034	0.0789474	0.0438596	0.2222222	0.1097724	0.7047619	0.5555556	0.5555556	0.3157895	0.1388889	0.1388889	0.9487179
bullish	AAPL	5	10	glm	0.7592593	0.5871795	0.0740741	0.0555556	0.3157895	0.2251656	0.7600000	0.7500000	0.7500000	0.2777778	0.2000000	0.2000000	0.9743590
bullish	AAPL	5	15	glm	0.6428571	0.4615385	0.0535714	0.0000000	NaN	-0.1001965	0.6792453	0.0000000	0.0000000	0.3035714	0.0000000	0.0000000	0.9230769
bullish	AAPL	5	20	glm	0.7037037	0.5076923	0.0555556	0.0185185	0.1111111	0.0204082	0.7254902	0.3333333	0.3333333	0.2777778	0.0666667	0.0666667	0.9487179
bullish	AAPL	5	1	glm - top layer	0.7982456	0.7254274	0.2192982	0.1666667	0.6229508	0.4912689	0.8089888	0.7600000	0.7600000	0.3157895	0.5277778	0.5277778	0.9230769
bullish	AAPL	5	5	glm - top layer	0.8245614	0.7745726	0.2631579	0.2017544	0.6969697	0.5749441	0.8452381	0.7666667	0.7666667	0.3157895	0.6388889	0.6388889	0.9102564
bullish	AAPL	5	10	glm - top layer	0.8796296	0.8141026	0.2129630	0.1851852	0.7547170	0.6767956	0.8823529	0.8695652	0.8695652	0.2777778	0.6666667	0.6666667	0.9615385
bullish	AAPL	5	15	glm - top layer	0.7767857	0.6904223	0.2053571	0.1428571	0.5614035	0.4190871	0.7977528	0.6956522	0.6956522	0.3035714	0.4705882	0.4705882	0.9102564
bullish	AAPL	5	20	glm - top layer	0.8518519	0.7948718	0.2407407	0.1851852	0.7142857	0.6149733	0.8780488	0.7692308	0.7692308	0.2777778	0.6666667	0.6666667	0.9230769
bullish	AAPL	5	1	nb	0.6842105	0.6570513	0.3684211	0.1842105	0.5384615	0.3006135	0.7916667	0.5000000	0.5000000	0.3157895	0.5833333	0.5833333	0.7307692
bullish	AAPL	5	5	nb	0.6754386	0.5683761	0.1842105	0.0877193	0.3508772	0.1540313	0.7204301	0.4761905	0.4761905	0.3157895	0.2777778	0.2777778	0.8589744
bullish	AAPL	5	10	nb	0.7129630	0.6166667	0.2314815	0.1111111	0.4363636	0.2459459	0.7831325	0.4800000	0.4800000	0.2777778	0.4000000	0.4000000	0.8333333
bullish	AAPL	5	15	nb	0.6785714	0.6033183	0.2678571	0.1250000	0.4375000	0.2137285	0.7560976	0.4666667	0.4666667	0.3035714	0.4117647	0.4117647	0.7948718
bullish	AAPL	5	20	nb	0.7592593	0.7205128	0.3148148	0.1759259	0.5937500	0.4236453	0.8513514	0.5588235	0.5588235	0.2777778	0.6333333	0.6333333	0.8076923
bullish	AAPL	5	1	nb - top layer	0.7719298	0.7136752	0.2631579	0.1754386	0.6060606	0.4474273	0.8095238	0.6666667	0.6666667	0.3157895	0.5555556	0.5555556	0.8717949
bullish	AAPL	5	5	nb - top layer	0.7894737	0.7564103	0.3157895	0.2105263	0.6666667	0.5128205	0.8461538	0.6666667	0.6666667	0.3157895	0.6666667	0.6666667	0.8461538
bullish	AAPL	5	10	nb - top layer	0.8611111	0.8012821	0.2314815	0.1851852	0.7272727	0.6351351	0.8795181	0.8000000	0.8000000	0.2777778	0.6666667	0.6666667	0.9358974
bullish	AAPL	5	15	nb - top layer	0.7410714	0.6481900	0.2053571	0.1250000	0.4912281	0.3261411	0.7752809	0.6086957	0.6086957	0.3035714	0.4117647	0.4117647	0.8846154
bullish	AAPL	5	20	nb - top layer	0.8518519	0.7948718	0.2407407	0.1851852	0.7142857	0.6149733	0.8780488	0.7692308	0.7692308	0.2777778	0.6666667	0.6666667	0.9230769

(VI) EVALUATION

In this final project, we focus on four key measures to evaluate the success of our models, i.e. accuracy, f1 score, precision, and recall (aka sensitivity). Overall, the model seems to be more accurate in pointint out a downward trend (“bearish”).

(VII) PREDICTION

Here’s the most important piece of this final project - prediction. Let’s first display the Moving Average chart. We look at the moving average of 5- and 20-trading windows. We can use this chart as a reference when reading our prediction chart.

The prediction chart displays the two trends, i.e. bearish, bullish, each is defined by the threshold that we set up earlier. The red line represents a downward (“bearish”) trend, whereas the green line represents the upward (“bullish”) trend. These lines move up and down along the trading dates. We should read it as, “on that specific trading date, we predict”t plus n trading day" for moving upward and crossing our threshold (above 85th percentile of our change in closing prices between t and t plus n day) would be XX%, whereas the probability of moving downward and crossing the threshold (below 15th percentile) would be YY%.

The red vertical dashed lines represent very strong signal of trending downward, whereas a green vertical dashed line represents the opposite upward trend. The vertical lines would only appear if there’s a consensus among the 5 ndays and they must cross above the 50% cut-off line. We would interpret that as a very strong signal that the coming future trading day(s) would see a very strong downward/upward trend. Looking at the prediction chart, we see red flags (bearish signals) for these dates, i.e. 2020-02-24, 2020-02-26, 2020-03-05, 2020-03-06, 2020-03-11. On the other hand, we see a green flag (bullish signal) for a specific date, i.e. 2020-03-24.

The model successfully captures these significant dates that were heavily impacted by the pandemic and its subsequent reactions from the global market. Basically, the market crashed on “02-24-2020” and the Dow plunged 1000 points on coronavirus fears and AAPL droped by almost 5% from the previous trading day (“02-21-2020”). The market went into complete turmoil in the week of “02-24-2020”. A month later, AAPL bounced back starting on “03-23-2020”. The prediction chart captured all these downard and upward moments in late March and throughout April. The prediction chart highly corresponds with the moving average chart and both charts display clear signal about when market will likely rise or crash.

CONCLUSION

Obviously there’s a lot of room for improvement, such as feature selection, precision, accuracy of the model, etc. Another piece of challenge is how to speed up the time of processing multiple stocks at the same time (running for 10 stocks would require almost an hour on my personal laptop with 32 GB of RAM). In addition, how do we decide an appropriate threshold that is useful or meaningful to us? It can vary from situation to situation, and obviously stock price movement depends on market volatility and other factors (like Elon Musk tweeted about TSLA price was too high in the morning and subsequently the market responded by dragging the price down by 12% before closing. All it took was a single tweet in a single trading day). After all, maybe all the hard work is not as easily interpretable or more insightful than a simple moving average chart!

APPENDIX

############################################################
### SET UP ###

# load packages
packages <- c('broom', 'caret', 'doParallel', 'e1071', 'foreach', 'glue', 'gridExtra', 'Hmisc', 'htmlwidgets', 'InformationValue', 'kableExtra', 'parallel', 'plotly', 'quantmod', 'sqldf', 'tictoc', 'tidyverse', 'TTR', 'wrapr')
pacman::p_load(char = packages)

# set environment
readRenviron("~/I/config/.env")

# set project home directory
PROJECT_HOME_DIRECTORY <- Sys.getenv("PROJECT_HOME_DIRECTORY")
setwd(PROJECT_HOME_DIRECTORY)

# get real time data
REAL_TIME = TRUE
ALPHA_VANTAGE_API <- Sys.getenv("ALPHA_VANTAGE_API")

### GET SYMBOLS ###

# stock symbol
stock <- c("AAPL")

# detect, use multicores
numCores <- parallel::detectCores()

# create a simple cluster on the local machine using all available threads
cl <- parallel::makeCluster(detectCores(), methods = FALSE)

# register our cluster
doParallel::registerDoParallel(cl)

# loop through a list of stock tickers - super fast! 5x faster than the traditional for-loop approach!!
symbols <- foreach::foreach(i = 1:length(stock), .errorhandling = 'remove') %dopar% { quantmod::getSymbols(stock[i]) } %>% unlist  # change .errorhandling = 'pass' to see error

# return a single list of xts objects from the valid symbols
if(REAL_TIME){
        
        xtsList <- vector(mode = "list", length = length(symbols))
        
        for(i in 1:length(symbols)){
                
                xtsList[[i]] <- quantmod::getSymbols(
                        
                        symbols[i], 
                        env = NULL,  # set env = NULL and that is equivalent to auto.assign = FALSE
                        src = "av",
                        periodicity = "daily", 
                        output.size = "full", 
                        adjusted = TRUE,
                        api.key = ALPHA_VANTAGE_API
                        
                )
                
                Sys.sleep(12)
                
        }
        
} else {
        
        xtsList <- foreach::foreach(i = 1:length(symbols)) %dopar% { quantmod::getSymbols(symbols[i], env = NULL, adjusted = TRUE) }  # set env = NULL and that is equivalent to auto.assign = FALSE        
        
}

# set names for xtsList
names(xtsList) <- symbols

############################################################
### CUSTOM FUNCTION ###
historicalTransformation <- function(xts,        
                                     l = 1,
                                     n = 1,
                                     originalSymbol = FALSE,
                                     transformOnly = FALSE
){
        
        # transform() function
        transformation <- function(xts, l, n, originalSymbol){
                
                # symbol
                symbol <- stringr::str_extract_all(names(xts), pattern = "^[[:alpha:]].*\\.") %>% unlist %>% stringr::str_to_lower(.) %>% unique %>% gsub("\\.", "", .)
                
                # rename columns
                names(xts) <- gsub("^[[:alpha:]].*\\.", "", names(xts)) %>% stringr::str_to_lower(.)
                
                # turn it into data.frame
                xts <- as.data.frame(xts)
                
                # insert columns
                xts <- xts %>%
                        dplyr::mutate(
                                # date
                                date = row.names(xts) %>% lubridate::ymd(.),
                                year = lubridate::year(date),
                                quarter = lubridate::quarter(date),
                                month = lubridate::month(date),
                                weekday = base::weekdays(date),
                                day = lubridate::day(date),
                                # Lag
                                op_l = quantmod::Lag(xts[, "open"], l),
                                hi_l = quantmod::Lag(xts[, "high"], l),
                                lo_l = quantmod::Lag(xts[, "low"], l),
                                cl_l = quantmod::Lag(xts[, "close"], l),
                                vol_l = quantmod::Lag(xts[, "volume"], l),
                                adj_l = quantmod::Lag(xts[, "adjusted"], l),
                                # Next
                                op_n = lead(xts[, "open"], n),
                                hi_n = lead(xts[, "high"], n),
                                lo_n = lead(xts[, "low"], n),
                                cl_n = lead(xts[, "close"], n),
                                vol_n = lead(xts[, "volume"], n),
                                adj_n = lead(xts[, "adjusted"], n),
                                # daily "Close - Open", "Hi - Low"
                                cl_op_diff = close - open,
                                hi_lo_diff = high - low,
                                # Lag "diff"
                                op_l_diff = open - op_l,
                                hi_l_diff = high - hi_l,
                                lo_l_diff = low - lo_l,
                                cl_l_diff = close - cl_l,
                                vol_l_diff = volume - vol_l,
                                adj_l_diff = adjusted - adj_l,
                                # Next "diff"
                                op_n_diff = op_n - open,
                                hi_n_diff = hi_n - high,
                                lo_n_diff = lo_n - low,
                                cl_n_diff = cl_n - close,
                                vol_n_diff = vol_n - volume,
                                adj_n_diff = adj_n - adjusted,
                                # oscillator - RSI, CCI                                
                                rsi_close = TTR::RSI(xts[, "close"]),
                                cci_close = TTR::CCI(xts[, "close"])
                        )
                
                # oscillator - MACD
                macd <- MACD(xts$close) %>% as.data.frame %>% dplyr::select(macd)
                signal <- MACD(xts$close) %>% as.data.frame %>% dplyr::select(signal)
                
                # add MACD
                xts <- xts %>%
                        dplyr::mutate(macd_close = macd$macd,
                                      signal_close = signal$signal,
                                      macd_close_diff = macd_close - signal_close) %>%
                        dplyr::select(date, year, quarter, month, weekday, day,
                                      open, high, low, close, volume, adjusted,
                                      rsi_close, cci_close, macd_close, signal_close, macd_close_diff,
                                      everything()) %>%
                        dplyr::arrange(date)
                
                # rename columns back with original symbol
                if(originalSymbol){names(xts) <- paste(symbol, names(xts), sep = "_")}
                
                # return xts
                return(xts)
                
        }
        
        # return a df object
        if(transformOnly){
                
                # return a transformed xts object
                transformObj <- transformation(xts, l, n, originalSymbol)
                return(transformObj)
                
        } else {
                
                # return the min and max of different variables associated with dates based on a transformed xts object
                SYMBOL <- transformation(xts, l, n, originalSymbol)
                
                minVector <- SYMBOL[complete.cases(SYMBOL), 6:ncol(SYMBOL)] %>% lapply(., min) %>% unlist
                maxVector <- SYMBOL[complete.cases(SYMBOL), 6:ncol(SYMBOL)] %>% lapply(., max) %>% unlist
                
                minList <- vector(mode = "list", length = length(minVector))
                for(i in 1:length(minList)){
                        minList[[i]] <- sqldf(sprintf("select date, %s from SYMBOL where floor(%s) = floor(%f)", names(minVector)[i], names(minVector)[i], minVector[i]))
                }
                
                maxList <- vector(mode = "list", length = length(maxVector))
                for(i in 1:length(maxList)){
                        maxList[[i]] <- sqldf(sprintf("select date, %s from SYMBOL where ceil(%s) = ceil(%f)", names(maxVector)[i], names(maxVector)[i], maxVector[i]))
                }
                
                minDf <- minList %>%
                        bind_rows() %>%
                        tidyr::gather(., key, value, -date) %>%
                        dplyr::filter(!is.na(value)) %>%
                        dplyr::mutate(type = "min") %>%
                        arrange(key, date)
                
                maxDf <- maxList %>%
                        bind_rows() %>%
                        tidyr::gather(., key, value, -date) %>%
                        dplyr::filter(!is.na(value)) %>%
                        dplyr::mutate(type = "max") %>%
                        arrange(key, date)
                
                historyObj <- bind_rows(minDf, maxDf) %>%
                        dplyr::select(., type, everything()) %>%
                        arrange(type, key, date)
                
                return(historyObj)
        }
        
}

############################################################
### SET PARAMETERS FOR MODELS ###

# trend
volatility = c("bullish", "bearish")

# target percentile threshold, e.g. top X% and bottom X% price change (cl_n_diff / close)
cl_percentile_threshold = 0.15

# xts subset
xts_data_subset = TRUE

# xts subset partition, i.e. number of most recent trading days
xts_data_subset_partition = 1000

# lag day(s)
l_period = c(1, 5, 10, 15, 20)

# next day(s)
n_period = c(1, 2, 3, 4, 5)

# baseTrainMethods
baseTrainMethods <- c("glm", "nb", "gbm")

# topTrainMethods
topTrainMethods <- c("glm", "nb", "gbm")

# tune length
tune.length = 3

# seed 
seed = 1234

# split data into train and test set by XX
size = 0.8

# Dependent Variable
DV <- "target"

# Features Selection Threshold
features_selection_threshold = .01

# dfEval, i.e. number of trading days for evaluation
dfEval_subset_partition = 90

# output
modelSummaryList <- vector(mode = "list")

evalList <- vector(mode = "list")

############################################################
### BUILD MODELS ###

# start timer
tic()

# start nested for() loop        
for(b in volatility){
        
        marketTrend = b
        
        for(l in l_period){
                
                lday = l
                
                for(i in n_period){
                        
                        nday = i
                        
                        for(j in stock){
                                
                                # get data
                                x <- xtsList[[j]]
                                
                                # get target threshold
                                if(xts_data_subset){
                                        y.t <- historicalTransformation(x, l = lday, n = nday, transformOnly = TRUE) %>%
                                                dplyr::mutate(percent_diff = cl_n_diff / close,
                                                              partition = (nrow(.)-1):0) %>%
                                                dplyr::filter(partition <= xts_data_subset_partition)
                                } else {
                                        y.t <- historicalTransformation(x, l = lday, n = nday, transformOnly = TRUE) %>%
                                                dplyr::mutate(percent_diff = cl_n_diff / close)
                                }
                                y <- y.t$percent_diff
                                y <- y[!is.na(y)]
                                z <- quantile(y, c(cl_percentile_threshold, 1 - cl_percentile_threshold))
                                
                                # features
                                features <- c(                                
                                        'rsi_close', 
                                        'cci_close',
                                        'macd_close', 
                                        'signal_close',
                                        'vol_n_diff',
                                        'op_l_percent_diff', 
                                        'hi_l_percent_diff', 
                                        'lo_l_percent_diff', 
                                        'cl_l_percent_diff', 
                                        'vol_l_percent_diff',    
                                        'cl_op_diff_by_op',
                                        'hi_lo_diff_by_op',
                                        'cl_op_diff_by_cl_l',
                                        'hi_lo_diff_by_cl_l',
                                        'cl_cl_l_ratio',
                                        'op_cl_l_ratio',
                                        'vol_cl_l_ratio',
                                        'vol_op_l_ratio',  
                                        'vol_diff_ratio',
                                        'sma_5',                        
                                        'sma_10',
                                        'sma_15'
                                )
                                
                                # transform, get DV, set partition
                                x.t <- historicalTransformation(x, l = lday, n = nday, transformOnly = TRUE) %>%
                                        dplyr::mutate(
                                                trend = marketTrend,
                                                symbol = j,                        
                                                l_period = lday,
                                                n_period = nday,
                                                op_l_percent_diff = ((open - op_l) / op_l) %>% as.vector,
                                                hi_l_percent_diff = ((high - hi_l) / hi_l) %>% as.vector,
                                                lo_l_percent_diff = ((low - lo_l) / lo_l) %>% as.vector,
                                                cl_l_percent_diff = ((close - cl_l) / cl_l) %>% as.vector,
                                                vol_l_percent_diff = ((volume - vol_l) / vol_l) %>% as.vector,
                                                cl_op_diff_by_op = ((close - open) / open) %>% as.vector,
                                                hi_lo_diff_by_op = ((high - low) / open) %>% as.vector,
                                                cl_op_diff_by_cl_l = ((close - open) / cl_l) %>% as.vector,
                                                hi_lo_diff_by_cl_l = ((high - low) / cl_l) %>% as.vector,
                                                cl_cl_l_ratio = (close / cl_l) %>% as.vector,
                                                op_cl_l_ratio = (open / cl_l) %>% as.vector, 
                                                vol_cl_l_ratio = (vol_l_percent_diff / cl_cl_l_ratio) %>% as.vector, 
                                                vol_op_l_ratio = (vol_l_percent_diff / op_cl_l_ratio) %>% as.vector,  
                                                vol_diff_ratio = ((vol_n_diff / vol_l_diff) %>% as.vector) %>% as.vector, 
                                                sma_5 = TTR::SMA(close, 5),                                
                                                sma_10 = TTR::SMA(close, 10),
                                                sma_15 = TTR::SMA(close, 15),
                                                target = dplyr::case_when(trend == "bullish" ~ ifelse( (cl_n_diff / close) > z[2], "Y", "N" ),
                                                                          trend == "bearish" ~ ifelse( (cl_n_diff / close) < z[1], "Y", "N" )) %>%
                                                        factor(., levels = c("Y", "N"), labels = c("Yes", "No")),            
                                                partition = (nrow(.)-1):0) %>%  # 0 is referred to the most recent (or last) trading day
                                        dplyr::select(trend, symbol, l_period, n_period, partition, date, year, quarter, month, weekday, day, everything()) %>%
                                        arrange(desc(partition))                
                                
                                # whether to subset and use only the most recent trading days or entire historical record
                                if(xts_data_subset){x.t <- x.t %>% dplyr::filter(partition <= xts_data_subset_partition)}
                                
                                # scale features
                                x.t.scaled <- x.t %>% dplyr::mutate_at(vars(features), scale)
                                
                                # remove any nan column
                                x.t.scaled <- x.t.scaled[, which(unlist(lapply(x.t.scaled, function(x) !any(is.nan(x)))))]
                                
                                # complete.cases
                                x.t.scaled <- x.t.scaled %>% complete.cases() %.>% x.t.scaled[., ]
                                
                                # new set of features
                                features <- features[features %in% names(x.t.scaled)]
                                
                                # feature selections by t.test        
                                t.test.list <- lapply(x.t.scaled[, features], function(x) t.test(x ~ x.t.scaled$target) %>% broom::tidy())
                                t.test.df <- t.test.list %>%
                                        dplyr::bind_rows() %>%
                                        dplyr::mutate(variables = features) %>%
                                        arrange(p.value)
                                
                                IV <- t.test.df$variables[t.test.df$p.value < features_selection_threshold]
                                if(length(IV) <=2){IV = unique(c(IV, 'rsi_close', 'cci_close', 'macd_close', 'signal_close'))}
                                
                                # subset features
                                x.t.scaled2 <- x.t.scaled %>% 
                                        dplyr::select(partition, date, all_of(IV), DV) %>%
                                        dplyr::filter(partition > dfEval_subset_partition)
                                
                                # evaluation set (include all variables and TODAY values)
                                dfEval <- x.t %>% 
                                        dplyr::filter(partition <= dfEval_subset_partition) %>%
                                        dplyr::mutate_at(vars(features), scale) %>%
                                        dplyr::select(partition, date, all_of(IV), DV) %>%                
                                        dplyr::mutate_at(IV, as.vector)
                                
                                # impute dataset
                                preProcValues <- caret::preProcess(dfEval[, IV], method = c("medianImpute"))
                                dfEval <- predict(preProcValues, dfEval) %>%
                                        dplyr::mutate_at(IV, as.vector)
                                
                                # oversampling target == "Yes" by 50%
                                x.t.scaled2_Yes <- x.t.scaled2 %>% dplyr::filter(target == "Yes")
                                x.t.scaled2_IndexUp <- sample(1:nrow(x.t.scaled2_Yes), size = floor(nrow(x.t.scaled2_Yes) * 1.5), replace = TRUE)
                                x.t.scaled2_Yes <- x.t.scaled2_Yes[x.t.scaled2_IndexUp, ]
                                
                                # undersampling target == "No" by 50%
                                x.t.scaled2_No <- x.t.scaled2 %>% dplyr::filter(target == "No")
                                x.t.scaled2_IndexDown <- sample(1:nrow(x.t.scaled2_No), size = floor(nrow(x.t.scaled2_No) * 0.5), replace = FALSE)
                                x.t.scaled2_No <- x.t.scaled2_No[x.t.scaled2_IndexDown, ]
                                
                                # put them together to come up with new x.t.scaled
                                x.t.scaled2 <- dplyr::bind_rows(x.t.scaled2_Yes, x.t.scaled2_No)
                                
                                # split data into train, test sets
                                set.seed(seed + nday + lday)
                                index <- caret::createDataPartition(x.t.scaled2$target, p = size, list = FALSE)
                                trainSet <- x.t.scaled2[index, ]
                                testSet <- x.t.scaled2[-index, ] %>% distinct 
                                
                                # trainSetControl
                                trainSet.control <- caret::trainControl(method = "cv", number = 10, savePredictions = "final", classProbs = TRUE)
                                
                                # train base layer
                                for(baseLayer in baseTrainMethods){
                                        
                                        # set parameters
                                        ml = baseLayer
                                        model = paste0("model_", ml)
                                        OOF_prediction = paste0("OOF_pred_", ml)
                                        prediction = paste0("pred_", ml)
                                        result = paste0("result_", ml)
                                        
                                        # model
                                        assign(bquote(.(model)), caret::train(trainSet[, IV], trainSet[, DV], 
                                                                              method = ml, 
                                                                              trControl = trainSet.control, 
                                                                              tuneLength = tune.length))
                                        
                                        # Out-Of-Fold probability predictions - trainSet    
                                        if(ml == "glm"){trainSet$OOF_pred_glm = eval(sym(model))$pred$Y[order(eval(sym(model))$pred$rowIndex)]}
                                        if(ml == "nb"){trainSet$OOF_pred_nb = eval(sym(model))$pred$Y[order(eval(sym(model))$pred$rowIndex)]}
                                        if(ml == "gbm"){trainSet$OOF_pred_gbm = eval(sym(model))$pred$Y[order(eval(sym(model))$pred$rowIndex)]}
                                        
                                        # Out-Of-Fold probability predictions - testSet 
                                        assign(bquote(.(OOF_prediction)), predict(eval(sym(model)), testSet[, IV], type = "prob")$Y)
                                        if(ml == "glm"){testSet$OOF_pred_glm = eval(sym(OOF_prediction))}
                                        if(ml == "nb"){testSet$OOF_pred_nb = eval(sym(OOF_prediction))}
                                        if(ml == "gbm"){testSet$OOF_pred_gbm = eval(sym(OOF_prediction))}
                                        
                                        # Out-Of-Fold probability predictions - dfEval 
                                        assign(bquote(.(OOF_prediction)), predict(eval(sym(model)), dfEval[, IV], type = "prob")$Y)
                                        if(ml == "glm"){dfEval$OOF_pred_glm = eval(sym(OOF_prediction))}
                                        if(ml == "nb"){dfEval$OOF_pred_nb = eval(sym(OOF_prediction))}
                                        if(ml == "gbm"){dfEval$OOF_pred_gbm = eval(sym(OOF_prediction))}
                                        
                                        # Y/N predictions for Confusion Matrix - testSet    
                                        assign(bquote(.(prediction)), predict(eval(sym(model)), testSet[, IV]))
                                        if(ml == "glm"){testSet$pred_glm = eval(sym(prediction))}
                                        if(ml == "nb"){testSet$pred_nb = eval(sym(prediction))}
                                        if(ml == "gbm"){testSet$pred_gbm = eval(sym(prediction))}
                                        
                                        # output
                                        assign(bquote(.(result)), broom::tidy(caret::confusionMatrix(testSet[, prediction], testSet$target)) %>%
                                                       dplyr::mutate(trend = marketTrend, symbol = j, n_period = nday, l_period = lday, trainMethod = ml) %>%
                                                       dplyr::select(trend, symbol, n_period, l_period, trainMethod, everything()))
                                        
                                        # store output into a list
                                        tempModelList <- list(eval(sym(result)))
                                        modelSummaryList <<- c(modelSummaryList, tempModelList)
                                        
                                }
                                
                                # train top layer
                                for(topLayer in topTrainMethods){
                                        
                                        # set parameters
                                        ml = topLayer
                                        model = paste0("model_", ml)        
                                        OOF_predictors_top = c("OOF_pred_glm", "OOF_pred_nb", "OOF_pred_gbm")
                                        OOF_prediction_top = paste0("OOF_pred_top_", ml)
                                        prediction_top = paste0("pred_top_", ml)
                                        result = paste0("result_top_", ml)
                                        
                                        # model
                                        assign(bquote(.(model)), caret::train(trainSet[, OOF_predictors_top], trainSet[, DV], 
                                                                              method = ml, 
                                                                              trControl = trainSet.control, 
                                                                              tuneLength = tune.length))
                                        
                                        # Out-Of-Fold probability predictions - testSet 
                                        assign(bquote(.(OOF_prediction_top)), predict(eval(sym(model)), testSet[, OOF_predictors_top], type = "prob")$Y)
                                        if(ml == "glm"){testSet$OOF_pred_top_glm = eval(sym(OOF_prediction_top))}
                                        if(ml == "nb"){testSet$OOF_pred_top_nb = eval(sym(OOF_prediction_top))}
                                        if(ml == "gbm"){testSet$OOF_pred_top_gbm = eval(sym(OOF_prediction_top))}
                                        
                                        # Out-Of-Fold probability predictions - dfEval 
                                        assign(bquote(.(OOF_prediction_top)), predict(eval(sym(model)), dfEval[, OOF_predictors_top], type = "prob")$Y)
                                        if(ml == "glm"){dfEval$OOF_pred_top_glm = eval(sym(OOF_prediction_top))}
                                        if(ml == "nb"){dfEval$OOF_pred_top_nb = eval(sym(OOF_prediction_top))}
                                        if(ml == "gbm"){dfEval$OOF_pred_top_gbm = eval(sym(OOF_prediction_top))}
                                        
                                        # Y/N predictions for Confusion Matrix - testSet    
                                        assign(bquote(.(prediction_top)), predict(eval(sym(model)), testSet[, OOF_predictors_top]))
                                        if(ml == "glm"){testSet$pred_top_glm = eval(sym(prediction_top))}
                                        if(ml == "nb"){testSet$pred_top_nb = eval(sym(prediction_top))}
                                        if(ml == "gbm"){testSet$pred_top_gbm = eval(sym(prediction_top))}
                                        
                                        # output
                                        assign(bquote(.(result)), broom::tidy(caret::confusionMatrix(testSet[, prediction_top], testSet$target)) %>%
                                                       dplyr::mutate(trend = marketTrend, symbol = j, n_period = nday, l_period = lday, trainMethod = paste0(ml, " - top layer")) %>%
                                                       dplyr::select(trend, symbol, n_period, l_period, trainMethod, everything()))
                                        
                                        # store output into a list
                                        tempModelList <- list(eval(sym(result)))
                                        modelSummaryList <<- c(modelSummaryList, tempModelList)
                                        
                                }
                                
                                # put together - final averaging
                                testSet <- testSet %>%
                                        dplyr::mutate(pred_final_avg = (OOF_pred_top_glm + OOF_pred_top_nb + OOF_pred_top_gbm) / length(topTrainMethods),
                                                      pred_final = ifelse(pred_final_avg > 0.5, "Y", "N") %>%
                                                              factor(., levels = c("Y", "N"), labels = c("Yes", "No")))
                                
                                dfEval <- dfEval %>%
                                        dplyr::mutate(pred_final_avg = (OOF_pred_top_glm + OOF_pred_top_nb + OOF_pred_top_gbm) / length(topTrainMethods),
                                                      pred_final = ifelse(pred_final_avg > 0.5, 1, 0),
                                                      trend = marketTrend, symbol = j, n_period = nday, l_period = lday) %>%
                                        dplyr::select(trend, symbol, n_period, l_period, date, target, pred_final_avg, pred_final)
                                
                                finalResult <- broom::tidy(caret::confusionMatrix(testSet$pred_final, testSet$target)) %>%
                                        dplyr::mutate(trend = marketTrend, symbol = j, n_period = nday, l_period = lday, trainMethod = "final - averaging") %>%
                                        dplyr::select(trend, symbol, n_period, l_period, trainMethod, everything())
                                
                                # store output into a list
                                tempModelList <- list(finalResult)
                                modelSummaryList <<- c(modelSummaryList, tempModelList)
                                
                                tempEvalList <- list(dfEval)
                                evalList <<- c(evalList, tempEvalList)
                                
                        }
                        
                }
                
        }
        
}

# stop timer
toc()

# consolidate lists into df
modelSummaryDf <- modelSummaryList %>% dplyr::bind_rows()
evalDf <- evalList %>% dplyr::bind_rows()

############################################################
### VISUALIZATION ###

###################################
### TECHNICAL ANALYSIS ###
x <- xts::xts(xtsList[[stock]])

quantmod::chartSeries( x["201801::202005"],
                       name = stock,
                       TA = c(addBBands(draw = 'bands'), 
                              addADX(), 
                              addMACD(), 
                              addRSI(), 
                              addCCI(), 
                              addCMF(), 
                              addBBands(draw = 'width'), 
                              addBBands(draw = 'percent'), 
                              addVo()) )

###################################
# density curve
tempObj <- vector(mode = "list", length = length(n_period))

for(i in n_period){
        
        tempDf <- xtsList[[stock]] %>% 
                historicalTransformation(transformOnly = TRUE, n = i) %>%
                dplyr::mutate(percent_change = (cl_n_diff / close),
                              t_period = i,
                              tday = dplyr::case_when(t_period == 1 ~ "t plus 1 day",
                                                      t_period == 2 ~ "t plus 2 days",
                                                      t_period == 3 ~ "t plus 3 days",
                                                      t_period == 4 ~ "t plus 4 days",
                                                      t_period == 5 ~ "t plus 5 days"),
                              index = nrow(.):1) %>%
                dplyr::filter(index <= xts_data_subset_partition) %>%
                dplyr::select(date, close, cl_n_diff, percent_change, tday)
        
        tempList <- list(tempDf)
        tempObj <- c(tempObj, tempList)
        
}

tempDf <- tempObj %>% dplyr::bind_rows()

# gather data frame
dfGather <- tempDf %>%
        dplyr::select(-date) %>%
        dplyr::mutate(tday = as.factor(tday)) %>%
        tidyr::gather(key, value, -tday) 

# plot close ($) by tday
densityCurveClose <- ggplot(dfGather %>% dplyr::filter(key == "close") %>% distinct, 
                            aes(value)) +
        geom_density() +
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "close") %>% distinct, 
                                    median), 
                   aes(xintercept = value),
                   linetype = "dashed",
                   col = "red") +
        theme(legend.position = "none") +
        labs(x = "Close ($)") +
        theme_light() +
        scale_x_continuous(labels = scales::dollar) +
        scale_y_continuous(labels = scales::percent) +
        ggtitle(paste0(stock, " distribution of Close ($) for past ", xts_data_subset_partition, " trading days"))

# plot cl_n_diff
densityCurveCl_N_diff <- ggplot(dfGather %>% dplyr::filter(key == "cl_n_diff"), 
                                aes(value, fill = tday)) +
        geom_density(alpha = 0.5) +
        # median
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "cl_n_diff"), 
                                    median), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dashed") +
        # cl_percentile_threshold
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "cl_n_diff"),
                                    quantile,
                                    probs = cl_percentile_threshold), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dotted") +
        # 1 - cl_percentile_threshold
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "cl_n_diff"),
                                    quantile,
                                    probs = 1 -cl_percentile_threshold), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dotted") +
        theme_minimal() +
        scale_x_continuous(labels = scales::dollar) +
        scale_y_continuous(labels = scales::percent) +
        theme(legend.position = "right") +
        labs(x = "Close ($) between t and t plus n day(s)") +
        ggtitle(paste0(stock, " distributions of difference between today (t) and future (t plus n trading day) Close ($) for past ", xts_data_subset_partition, " trading days")) +
        facet_wrap(~tday, nrow = 5)

# plot percent_change
densityCurveCl_percent_change <- ggplot(dfGather %>% dplyr::filter(key == "percent_change"), 
                                        aes(value, fill = tday)) +
        geom_density(alpha = 0.5) +
        # median
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "percent_change"), 
                                    median), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dashed") +
        # cl_percentile_threshold
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "percent_change"),
                                    quantile,
                                    probs = cl_percentile_threshold), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dotted") +
        # 1 - cl_percentile_threshold
        geom_vline(data = aggregate(value ~ tday, 
                                    dfGather %>% dplyr::filter(key == "percent_change"),
                                    quantile,
                                    probs = 1 -cl_percentile_threshold), 
                   aes(xintercept = value,
                       color = tday),
                   linetype = "dotted") +
        theme_bw() +
        scale_x_continuous(labels = scales::percent) +
        scale_y_continuous(labels = scales::percent) +
        theme(legend.position = "right") +
        labs(x = "% diff between t and t plus n day(s)") +
        ggtitle(paste0(stock, " distributions of change in percentage between today (t) and future (t plus n trading day) Close ($) for past ", xts_data_subset_partition, " trading days")) +
        facet_wrap(~tday, nrow = 5)

###################################
# evaluation chart
# data transformation for model evaluation

# spread table
modelSummarySpread <- modelSummaryDf %>% 
        dplyr::select(trend, symbol, n_period, l_period, trainMethod, term, estimate) %>%
        tidyr::spread(term, estimate) %>%
        arrange(trend, symbol, n_period, trainMethod)

# get measures
modelSummaryMeasure <- modelSummaryDf %>%
        dplyr::filter(term %in% c("recall", "precision", "f1") & trainMethod == "final - averaging") %>%
        dplyr::select(trend, symbol, n_period, l_period, measure = term, estimate) %>%
        dplyr::mutate(lday = dplyr::case_when(l_period == 1 ~ "lag 1 day",
                                              l_period == 5 ~ "lag 5 days",
                                              l_period == 10 ~ "lag 10 days",
                                              l_period == 15 ~ "lag 15 days",
                                              l_period == 20 ~ "lag 20 days") %>%
                              factor(., levels = c("lag 1 day", "lag 5 days", "lag 10 days", "lag 15 days", "lag 20 days")))

# get accuracy from eval set
accuracyDf <- evalDf %>%
        dplyr::mutate(flag = dplyr::case_when(pred_final == 0 ~ "No", TRUE ~ "Yes"),
                      accuracy = dplyr::case_when(target == flag ~ 1, TRUE ~ 0)) %>%
        dplyr::filter(!is.na(target)) %>%
        group_by(trend, symbol, n_period, l_period) %>%
        summarise(accuracy = mean(accuracy)) %>%
        dplyr::mutate(measure = "accuracy",
                      lday = dplyr::case_when(l_period == 1 ~ "lag 1 day",
                                              l_period == 5 ~ "lag 5 days",
                                              l_period == 10 ~ "lag 10 days",
                                              l_period == 15 ~ "lag 15 days",
                                              l_period == 20 ~ "lag 20 days") %>%
                              factor(., levels = c("lag 1 day", "lag 5 days", "lag 10 days", "lag 15 days", "lag 20 days"))) %>%
        dplyr::select(trend, symbol, n_period, l_period, measure, estimate = accuracy, lday) %>%
        arrange(trend, symbol, n_period, l_period)

# combine the key measures of model summary together
keyMeasures <- dplyr::bind_rows(modelSummaryMeasure, accuracyDf) %>%
        arrange(trend, symbol, n_period, l_period, measure)

# plot evaluation chart
evalChart <- keyMeasures %>%
        dplyr::filter(symbol == stock) %>%
        ggplot(aes(n_period, estimate, fill = factor(trend))) +
        geom_bar(stat = "identity", position = "dodge2") +
        geom_hline(yintercept = 0.5, linetype = "dashed", col = "black") +
        scale_y_continuous(labels = scales::percent, limits = c(0, 1)) +
        facet_grid(measure ~ lday) +
        labs(x = "Prediction for t plus n trading day(s)", y = "", fill = "trend") +
        theme_light() +
        theme(legend.position = "bottom") +
        ggtitle(paste0(stock, " Model Evaluation ", min(evalDf$date), " : ", max(evalDf$date)))

###################################
# MA charts
MA_chart <- xtsList[[stock]] %>% 
        historicalTransformation(transformOnly = TRUE) %>%
        dplyr::mutate(`5 day MA` = TTR::SMA(close, 5),
                      `20 day MA` = TTR::SMA(close, 20),
                      index = nrow(.):1) %>%
        dplyr::filter(index <= dfEval_subset_partition) %>%
        dplyr::select(date, close, `5 day MA`, `20 day MA`) %>%
        tidyr::gather(key, value, -date) %>%
        ggplot(aes(date, value)) + 
        geom_line(aes(col = key)) +
        labs(x = "", y = "") +
        scale_x_date(labels = scales::date_format("%Y-%m-%d"), date_breaks = "3 day") +
        scale_color_manual(values = c("black", "blue", "red")) +
        theme_minimal() +
        theme(legend.position = "top", 
              plot.title = element_text(hjust = 0.5),
              axis.text.x = element_text(hjust = 1, angle = 60),
              legend.title = element_blank()) +
        scale_y_continuous(labels = scales::dollar) +
        guides(col = guide_legend(reverse = TRUE)) +
        ggtitle(paste0(stock, " Close ($) Moving Average(s) for past ", dfEval_subset_partition, " trading days"))


###################################
# prediction chart

# data transformation for prediction
predDf <- evalDf %>%
        group_by(trend, symbol, n_period, date) %>%
        summarise(prediction = round(mean(pred_final_avg), 3)) %>% 
        dplyr::mutate(tday = dplyr::case_when(n_period == 1 ~ "t plus 1 day",
                                              n_period == 2 ~ "t plus 2 days",
                                              n_period == 3 ~ "t plus 3 days",
                                              n_period == 4 ~ "t plus 4 days",
                                              n_period == 5 ~ "t plus 5 days")) %>%
        ungroup %>%
        dplyr::select(-n_period)

# get signal dates
signalDate <- predDf %>%
        dplyr::select(trend, symbol, date, prediction, tday) %>%
        tidyr::spread(tday, prediction) %>%
        dplyr::mutate(`t plus 1 day` = dplyr::case_when(`t plus 1 day` > 0.5 ~ 1, TRUE ~ 0),
                      `t plus 2 days` = dplyr::case_when(`t plus 2 days` > 0.5 ~ 1, TRUE ~ 0),
                      `t plus 3 days` = dplyr::case_when(`t plus 3 days` > 0.5 ~ 1, TRUE ~ 0),
                      `t plus 4 days` = dplyr::case_when(`t plus 4 days` > 0.5 ~ 1, TRUE ~ 0),
                      `t plus 5 days` = dplyr::case_when(`t plus 5 days` > 0.5 ~ 1, TRUE ~ 0),
                      flag = `t plus 1 day` + `t plus 2 days` + `t plus 3 days` + `t plus 4 days` + `t plus 5 days`) %>%
        dplyr::select(trend, symbol, date, flag) %>%
        tidyr::spread(trend, flag) %>%
        dplyr::mutate(signal = dplyr::case_when(bearish <= 2 & bullish == 5 ~ 1,
                                                bearish == 5 & bullish <= 2 ~ 1,
                                                TRUE ~ 0)) %>%
        dplyr::filter(signal == 1) %>% 
        dplyr::select(-signal) %>%
        tidyr::gather(trend, value, -symbol, -date) %>%
        dplyr::filter(value == 5) %>%
        dplyr::select(-value)

# plot prediction chart
predChart <- predDf %>%
        dplyr::filter(symbol == stock) %>%
        ggplot(aes(x = date, y = prediction, col = trend)) +
        geom_line() +
        geom_hline(yintercept = 0.5, linetype = "dashed", col = "black") +
        geom_vline(aes(xintercept = date, col = trend), 
                   data = signalDate %>% dplyr::filter(symbol == stock),
                   linetype = "dashed") +
        scale_y_continuous(labels = scales::percent, limits = c(0, 1)) +
        scale_x_date(labels = scales::date_format("%Y-%m-%d"), 
                     date_breaks = "3 day", 
                     limits = c(min(predDf$date), max(predDf$date))) +
        facet_wrap(~ tday, nrow = 5) +
        theme_minimal() +
        theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
        theme(legend.position = "bottom") +
        labs(x = "", y = "") +
        ggtitle(paste0(stock, " Trend Prediction ", min(predDf$date), " : ", max(predDf$date)))

###################################
### PLOTLY ###

densityCurveClose %>% ggplotly()

densityCurveCl_N_diff %>% ggplotly()

densityCurveCl_percent_change %>% ggplotly()

evalChart %>% ggplotly()

MA_chart %>% ggplotly()

predChart  # vertical dashlines would disappear if we convert it into a plotly object, thus not recommend

###################################
### END ###
# stop the cluster
parallel::stopCluster(cl)