Time Series Forecasting - Modeltime and Slidy

Tyler Otto

February 20, 2022
• c - table of contents (bottom left)
• k - please disable click to next slide to ensure interactivity works
• use arrow keys to navigate

Overview

5 types of models will be built and the best fit version of each model will be chosen an the champion for each type. The example dataset is the popular Superstore dataset.

        - Linear Regression Model
        - ARIMA (Autoregressive Integrated Moving Average)
        - GLMNET (Lasso and Elastic-Net Regularized Generalized Linear Models)
        - Earth/MARS (Multivariate Adaptive Regression Spline)
        - ETS (Exponential Smoothing)

Models are chosen based on aggregating the accuracy measures and choosing the one with the least errors
Modeltime documentation: https://business-science.github.io/modeltime/

Time Series Testing - Consumer Segment Sales

Accuracy Table
.model_id	.model_desc	.type	mae	mape	mase	smape	rmse	ensemble_score
1	ARIMA(0,0,0)(1,1,0)[12] WITH DRIFT W/ XGBOOST ERRORS	Test	110.70	21.96	0.60	23.78	135.21	20
2	LM	Test	96.96	18.34	0.53	19.45	128.29	14
3	ETS(M,N,M)	Test	124.25	22.05	0.68	25.80	152.38	25
4	EARTH	Test	63.78	15.11	0.35	13.52	90.42	5
5	GLMNET	Test	87.57	18.42	0.48	17.99	112.42	11

Time Series Forecast - Consumer Segment Sales

sales_month	2018_UPDATE: ARIMA(0,1,1)	2018_LM	2018_UPDATE: ETS(M,A,M)	2018_EARTH	2018_GLMNET
Jan	487.0	341.0	300.0	598	263.0
Feb	400.0	284.0	202.0	455	224.0
Mar	606.0	470.0	476.0	640	404.0
Apr	596.0	449.0	459.0	662	474.0
May	720.0	515.0	574.0	684	473.0
Jun	688.0	513.0	580.0	706	542.0
Jul	658.0	499.0	532.0	727	525.0
Aug	660.0	522.0	581.0	749	554.0
Sep	1,043.0	837.0	1,075.0	1,107	857.0
Oct	657.0	490.0	525.0	793	420.0
Nov	1,083.0	908.0	1,216.0	1,207	841.0
Dec	1,071.0	878.0	1,141.0	1,191	898.0
Total	8,674.0	6,711.0	7,666.0	9,524	6,482.0

Time Series Testing - Corporate Segment Sales

Accuracy Table
.model_id	.model_desc	.type	mae	mape	mase	smape	rmse	ensemble_score
1	ARIMA(0,0,0)(1,1,0)[12] WITH DRIFT W/ XGBOOST ERRORS	Test	42.50	15.95	0.46	15.86	48.84	5
2	LM	Test	49.81	17.00	0.54	17.60	58.88	10
3	ETS(M,N,M)	Test	70.80	21.65	0.76	24.95	85.96	20
4	EARTH	Test	121.00	55.43	1.30	38.07	133.08	25
5	GLMNET	Test	55.42	17.50	0.60	18.42	65.73	15

Time Series Forecast - Corporate Segment Sales

sales_month	2018_ARIMA(0,0,0)(1,1,0)[	2018_LM	2018_ETS(M,N,M)	2018_EARTH	2018_GLMNET
Jan	217	207.0	152.0	311.0	189.0
Feb	181	185.0	111.0	221.0	212.0
Mar	289	302.0	264.0	322.0	326.0
Apr	242	281.0	269.0	329.0	282.0
May	357	322.0	287.0	335.0	313.0
Jun	304	285.0	240.0	341.0	294.0
Jul	340	318.0	280.0	347.0	320.0
Aug	320	315.0	261.0	353.0	306.0
Sep	516	472.0	449.0	528.0	452.0
Oct	393	378.0	336.0	438.0	366.0
Nov	557	543.0	537.0	604.0	566.0
Dec	599	531.0	517.0	593.0	511.0
Total	4,322	4,144.0	3,708.0	4,726.0	4,144.0

Time Series Testing - Home Office Segment Sales

Accuracy Table
.model_id	.model_desc	.type	mae	mape	mase	smape	rmse	ensemble_score
1	ARIMA(0,1,1)(0,1,1)[12] W/ XGBOOST ERRORS	Test	43.37	20.54	0.66	22.05	51.37	10
2	LM	Test	46.10	21.49	0.70	23.87	55.56	15
3	ETS(A,N,A)	Test	48.45	21.52	0.73	24.83	62.48	21
4	EARTH	Test	41.39	19.73	0.63	20.23	49.40	5
5	GLMNET	Test	48.85	23.13	0.74	25.81	59.36	24

Time Series Forecast - Home Office Segment Sales

sales_month	2018_UPDATE: ARIMA(0,1,1)	2018_LM	2018_UPDATE: ETS(M,N,A)	2018_EARTH	2018_GLMNET
Jan	207.0	164.0	159.0	207.0	153.0
Feb	172.0	140.0	135.0	151.0	135.0
Mar	266.0	212.0	202.0	214.0	204.0
Apr	268.0	225.0	217.0	218.0	216.0
May	194.0	203.0	203.0	222.0	208.0
Jun	256.0	216.0	207.0	225.0	213.0
Jul	207.0	202.0	193.0	229.0	194.0
Aug	199.0	202.0	196.0	232.0	217.0
Sep	339.0	300.0	274.0	317.0	311.0
Oct	337.0	251.0	222.0	240.0	238.0
Nov	436.0	336.0	286.0	355.0	328.0
Dec	256.0	289.0	274.0	309.0	300.0
Total	3,139.0	2,745.0	2,574.0	2,926.0	2,723.0