siva final project
siva
2023-10-12
R Markdown
This is an R Markdown document. Markdown is a simple formatting syntax for authoring HTML, PDF, and MS Word documents. For more details on using R Markdown see http://rmarkdown.rstudio.com.
When you click the Knit button a document will be generated that includes both content as well as the output of any embedded R code chunks within the document. You can embed an R code chunk like this:
Loading library and reading the data.
install.packages("moments")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'moments' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packageslibrary(readr)
seattle_weather <- read_csv("seattle-weather.csv")## Rows: 1461 Columns: 6
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (2): date, weather
## dbl (4): precipitation, temp_max, temp_min, wind
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Checking for missing values.
missing_data <- is.na(seattle_weather)
# Count the missing values for each variable
missing_counts <- colSums(missing_data)
# Print the number of missing values for each variable
print("Missing values in each variable:")## [1] "Missing values in each variable:"print(missing_counts)## date precipitation temp_max temp_min wind
## 0 0 0 0 0
## weather
## 0# Summarize the total number of missing values in the dataset
total_missing_values <- sum(missing_counts)
print(paste("Total missing values in the dataset:", total_missing_values))## [1] "Total missing values in the dataset: 0"from the above results i can say there are no missing values in the data.
changing date format into standard format.
seattle_weather$date <- as.Date(seattle_weather$date, format = "%d/%m/%Y")
# Where:
# - 'seattle_weather' is the name of your data frame.
# - 'date' is the name of the column you want to convert.
# - 'format' specifies the current date format ("%d/%m/%Y" for "01/01/2012").
# Print the first few rows to verify the conversion
head(seattle_weather)## # A tibble: 6 × 6
## date precipitation temp_max temp_min wind weather
## <date> <dbl> <dbl> <dbl> <dbl> <chr>
## 1 2012-01-01 0 12.8 5 4.7 drizzle
## 2 2012-01-02 10.9 10.6 2.8 4.5 rain
## 3 2012-01-03 0.8 11.7 7.2 2.3 rain
## 4 2012-01-04 20.3 12.2 5.6 4.7 rain
## 5 2012-01-05 1.3 8.9 2.8 6.1 rain
## 6 2012-01-06 2.5 4.4 2.2 2.2 rainCreating Histogram for temp_min.
library(ggplot2)## Warning: package 'ggplot2' was built under R version 4.3.1# Create a histogram for temperaturemin
ggplot(seattle_weather, aes(x = temp_min)) +
geom_histogram(binwidth = 2, fill = "blue", color = "black") +
labs(title = "Distribution of Minimum Temperature", x = "Temperature (°C)", y = "Frequency")
summary(seattle_weather)## date precipitation temp_max temp_min
## Min. :2012-01-01 Min. : 0.000 Min. :-1.60 Min. :-7.100
## 1st Qu.:2012-12-31 1st Qu.: 0.000 1st Qu.:10.60 1st Qu.: 4.400
## Median :2013-12-31 Median : 0.000 Median :15.60 Median : 8.300
## Mean :2013-12-31 Mean : 3.029 Mean :16.44 Mean : 8.235
## 3rd Qu.:2014-12-31 3rd Qu.: 2.800 3rd Qu.:22.20 3rd Qu.:12.200
## Max. :2015-12-31 Max. :55.900 Max. :35.60 Max. :18.300
## wind weather
## Min. :0.400 Length:1461
## 1st Qu.:2.200 Class :character
## Median :3.000 Mode :character
## Mean :3.241
## 3rd Qu.:4.000
## Max. :9.500Results from summary:
the data set period is from jan 1,2012 to dec 31,2015. for precipitation min is 0 and max is 55.90 and mean is 3.029. we can see right skew by comparing mean and max precipitation.
for temp max temp is 35.60 and min is -1.60 and mean is 16.44. there is little right skew in the data.
for temp_min min is -7.100 and max is 18.30 and mean is 8.235. right skew.
same way wind speed vary from 0.4 m/s to 9.5 m/s, with the mean of 3.241.
finding for skewness. weather data is normal or skewed. If the data is skewed we need to convert to normal.
# Install and load the 'moments' package if you haven't already
install.packages("moments")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'moments' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packageslibrary(moments)
# Assuming you have a data frame named 'seattle_weather'
# Calculate skewness for each numeric variable separately
skewness_temp_max <- skewness(seattle_weather$temp_max, na.rm = TRUE)
skewness_temp_min <- skewness(seattle_weather$temp_min, na.rm = TRUE)
skewness_wind <- skewness(seattle_weather$wind, na.rm = TRUE)
skewness_precipitation <- skewness(seattle_weather$precipitation, na.rm = TRUE)
# Print the skewness values
cat("Skewness Value for 'temp_max':", skewness_temp_max, "\n")## Skewness Value for 'temp_max': 0.2806415cat("Skewness Value for 'temp_min':", skewness_temp_min, "\n")## Skewness Value for 'temp_min': -0.2492024cat("Skewness Value for 'wind':", skewness_wind, "\n")## Skewness Value for 'wind': 0.8907518cat("Skewness Value for 'precipitation':", skewness_precipitation, "\n")## Skewness Value for 'precipitation': 3.502043from the above result the skewness for all is in between -1 to +1, except for precipitation. for precipitation skewness is greater than 1. So i need to perform log function on precipitation to convert into normal data.
performing log function on precipitation.
# Add a small constant before log transformation
small_constant <- 0.001
seattle_weather$log_precipitation <- log(seattle_weather$precipitation + small_constant)Again finding skewness for logerthemic precipitation.
install.packages("moments")## Warning: package 'moments' is in use and will not be installedlibrary(moments)
# Calculate skewness for the log-transformed precipitation data
skewness_precipitation <- skewness(seattle_weather$log_precipitation, na.rm = TRUE)
# Print the skewness value for log-transformed precipitation
cat("Skewness Value for Log-Transformed Precipitation Data:", skewness_precipitation, "\n")## Skewness Value for Log-Transformed Precipitation Data: 0.4271911from the above result the skewness is -1 to +1. we are good to go.
seattle_weather## # A tibble: 1,461 × 7
## date precipitation temp_max temp_min wind weather log_precipitation
## <date> <dbl> <dbl> <dbl> <dbl> <chr> <dbl>
## 1 2012-01-01 0 12.8 5 4.7 drizzle -6.91
## 2 2012-01-02 10.9 10.6 2.8 4.5 rain 2.39
## 3 2012-01-03 0.8 11.7 7.2 2.3 rain -0.222
## 4 2012-01-04 20.3 12.2 5.6 4.7 rain 3.01
## 5 2012-01-05 1.3 8.9 2.8 6.1 rain 0.263
## 6 2012-01-06 2.5 4.4 2.2 2.2 rain 0.917
## 7 2012-01-07 0 7.2 2.8 2.3 rain -6.91
## 8 2012-01-08 0 10 2.8 2 sun -6.91
## 9 2012-01-09 4.3 9.4 5 3.4 rain 1.46
## 10 2012-01-10 1 6.1 0.6 3.4 rain 0.00100
## # ℹ 1,451 more rowsi am performing standardisation on the data. so that all input variables comes to single scale with mean 0 and standard deviation.
# Standardization (Z-Score Normalization)
seattle_weather_standardized <- seattle_weather
# Define the numeric columns you want to standardize
numeric_cols <- c("temp_max", "temp_min", "wind", "log_precipitation")
# Apply Z-Score normalization
seattle_weather_standardized[, numeric_cols] <- scale(seattle_weather_standardized[, numeric_cols])
# View the first few rows of the standardized data
head(seattle_weather_standardized)## # A tibble: 6 × 7
## date precipitation temp_max temp_min wind weather log_precipitation
## <date> <dbl> <dbl> <dbl> <dbl> <chr> <dbl>
## 1 2012-01-01 0 -0.495 -0.644 1.01 drizzle -0.842
## 2 2012-01-02 10.9 -0.794 -1.08 0.876 rain 1.42
## 3 2012-01-03 0.8 -0.645 -0.206 -0.655 rain 0.783
## 4 2012-01-04 20.3 -0.577 -0.525 1.01 rain 1.57
## 5 2012-01-05 1.3 -1.03 -1.08 1.99 rain 0.901
## 6 2012-01-06 2.5 -1.64 -1.20 -0.724 rain 1.06splitting the data. I am using constant random seed to get unique result whenever i run the code.
library(rsample)## Warning: package 'rsample' was built under R version 4.3.1set.seed(20230712)
train_test_split <- initial_split(seattle_weather_standardized, prop = .80)
data_train <- training(train_test_split)
data_test <- testing(train_test_split)data_test## # A tibble: 293 × 7
## date precipitation temp_max temp_min wind weather log_precipitation
## <date> <dbl> <dbl> <dbl> <dbl> <chr> <dbl>
## 1 2012-01-02 10.9 -0.794 -1.08 0.876 rain 1.42
## 2 2012-01-14 4.1 -1.64 -1.52 1.43 snow 1.18
## 3 2012-01-17 8.1 -1.79 -1.64 1.64 snow 1.35
## 4 2012-01-18 19.8 -2.24 -2.20 1.22 snow 1.56
## 5 2012-01-19 15.2 -2.39 -2.20 -1.14 snow 1.50
## 6 2012-01-26 4.8 -1.03 -1.42 1.08 rain 1.22
## 7 2012-01-29 27.7 -0.958 -0.863 0.876 rain 1.64
## 8 2012-02-08 2.8 -0.876 -0.644 -0.376 rain 1.09
## 9 2012-02-15 0 -1.26 -1.52 -1.00 drizzle -0.842
## 10 2012-02-20 3 -1.18 -1.30 -0.237 rain 1.10
## # ℹ 283 more rowsdata_train## # A tibble: 1,168 × 7
## date precipitation temp_max temp_min wind weather log_precipitation
## <date> <dbl> <dbl> <dbl> <dbl> <chr> <dbl>
## 1 2012-06-04 1.3 -0.495 0.132 -0.0982 rain 0.901
## 2 2013-07-03 0 1.31 1.69 -0.0286 sun -0.842
## 3 2013-05-20 0 0.403 0.232 -1.00 sun -0.842
## 4 2014-12-10 13 0.335 0.351 2.41 rain 1.46
## 5 2015-12-02 2.5 -0.794 -0.763 1.22 rain 1.06
## 6 2014-01-25 0 -0.577 -1.42 -1.70 sun -0.842
## 7 2015-05-24 0 0.185 0.570 -0.376 sun -0.842
## 8 2013-12-21 5.6 -1.03 -0.525 -0.655 rain 1.26
## 9 2014-03-13 0.5 -0.345 -0.644 -0.515 rain 0.669
## 10 2014-03-17 0.3 -0.876 -1.08 -0.0286 rain 0.545
## # ℹ 1,158 more rowscreating recipe for the model. i am considering temp_max, temp_min,wind,log_precipitation as feature variables.
install.packages("magrittr")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'magrittr' successfully unpacked and MD5 sums checked## Warning: cannot remove prior installation of package 'magrittr'## Warning in file.copy(savedcopy, lib, recursive = TRUE): problem copying
## C:\Users\sivak\AppData\Local\R\win-library\4.3\00LOCK\magrittr\libs\x64\magrittr.dll
## to
## C:\Users\sivak\AppData\Local\R\win-library\4.3\magrittr\libs\x64\magrittr.dll:
## Permission denied## Warning: restored 'magrittr'##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packagesoptions(warn.conflicts = FALSE)
library(recipes)## Warning: package 'recipes' was built under R version 4.3.1## Loading required package: dplyr##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union##
## Attaching package: 'recipes'## The following object is masked from 'package:stats':
##
## steplibrary(magrittr)library(workflows)## Warning: package 'workflows' was built under R version 4.3.1library(parsnip)## Warning: package 'parsnip' was built under R version 4.3.1my_rec <- recipe(weather ~ temp_max+ temp_min + wind + log_precipitation , data = data_train)
my_rec## ## ── Recipe ──────────────────────────────────────────────────────────────────────## ## ── Inputs## Number of variables by role## outcome: 1
## predictor: 4from this we can say there are 4 predictor or feature variables available to predict the weather condition.
# Assuming 'data_train' is your data frame
seattle_weather$weather <- as.factor(seattle_weather$weather)levels(seattle_weather$weather)## [1] "drizzle" "fog" "rain" "snow" "sun"Installing nnet packages and constructing the multinomial logistic regression model because the output variable is not binary. there are 4 classes in the output.
install.packages("nnet") # Install the 'nnet' package## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'nnet' successfully unpacked and MD5 sums checked## Warning: cannot remove prior installation of package 'nnet'## Warning in file.copy(savedcopy, lib, recursive = TRUE): problem copying
## C:\Users\sivak\AppData\Local\R\win-library\4.3\00LOCK\nnet\libs\x64\nnet.dll to
## C:\Users\sivak\AppData\Local\R\win-library\4.3\nnet\libs\x64\nnet.dll:
## Permission denied## Warning: restored 'nnet'##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packageslibrary(nnet) # Load the 'nnet' package## Warning: package 'nnet' was built under R version 4.3.1# Install and load the 'nnet' package
install.packages("nnet")## Warning: package 'nnet' is in use and will not be installedlibrary(nnet)
# Fit the multinomial logistic regression model
multinom_model <- multinom(weather ~ temp_max + temp_min + wind + log_precipitation, data = data_train)## # weights: 30 (20 variable)
## initial value 1879.823482
## iter 10 value 695.743897
## iter 20 value 529.159979
## iter 30 value 527.556250
## iter 40 value 527.365101
## iter 50 value 527.313081
## iter 60 value 527.302171
## iter 70 value 527.292290
## final value 527.292018
## converged# Summary of the model
summary(multinom_model)## Call:
## multinom(formula = weather ~ temp_max + temp_min + wind + log_precipitation,
## data = data_train)
##
## Coefficients:
## (Intercept) temp_max temp_min wind log_precipitation
## fog 3.035856 0.01800714 0.2543537 0.1913660 2.7186955
## rain 13.163772 -0.58936404 0.5849886 0.9962782 15.5386525
## snow -2.715052 -3.86803531 -3.3271336 1.5239656 19.1788878
## sun 3.045500 1.07381872 -0.5795547 0.7517946 0.4814626
##
## Std. Errors:
## (Intercept) temp_max temp_min wind log_precipitation
## fog 0.1489662 0.4527075 0.4196636 0.2786007 0.1254845
## rain 0.5015885 0.5712987 0.5181402 0.2932140 0.5801325
## snow 2.9020567 1.6271192 1.1637404 0.3893581 1.5583548
## sun 0.1278680 0.3892005 0.3579674 0.2375716 0.1077126
##
## Residual Deviance: 1054.584
## AIC: 1094.584fog
Intercept: After accounting for wind, temperature min, temperature max, and log precipitation, the log-odds of having foggy weather are around 3.036 when all other predictors are zero. temp_max: Keeping all other variables equal, a one-unit increase in temp_max is correlated with a very little increase in the log-odds of the weather being foggy (0.018). temp_min: Keeping all other variables equal, a one-unit increase in temp_min is correlated with a little increase in the log-odds of the weather being foggy (0.254). wind: Keeping all other factors equal, a one-unit increase in wind is correlated with a little increase in the log-odds of the weather being foggy (0.191).
log_precipitation: Keeping other factors equal, a one-unit increase in log_precipitation corresponds to a 2.72-fold increase in the likelihood of cloudy weather.
Rainfall:
Intercept: At 0 for all other predictors, the log-odds of wet weather are around 13.164. temp_max: Keeping all other factors equal, a one-unit rise in temp_max is correlated with a -0.589 drop in the log-odds of wet weather. temp_min: Keeping all other variables equal, a one-unit increase in temp_min is correlated with an increase in the log-odds of it raining (0.585). wind: Keeping all other factors equal, a one-unit increase in wind is correlated with a one-unit increase in the log-odds of wet weather (0.996). log_precipitation: Keeping other factors equal, a one-unit increase in log_precipitation corresponds to a 15.54-fold increase in the likelihood of wet weather.
Snowfall:
Intercept: At zero for all other predictors, the log-odds of snowy weather are around -2.715. temp_max: Keeping other factors equal, a one-unit rise in temp_max is correlated with a considerable drop in the log-odds of the weather being snowy (-3.868). temp_min: When all other factors are held constant, a one-unit rise in temp_min is correlated with a substantial drop in the log-odds of the weather being snowy (-3.327). wind: Keeping all factors equal, a one-unit increase in wind is correlated with a one-unit increase in the log-odds of snowy weather (1.524). log_precipitation: Keeping all other factors equal, a one-unit increase in log_precipitation corresponds to a 19.18-fold increase in the likelihood of snowy weather.
Sun:
Intercept: There is an approximate 3.045 log-odds chance of sunny weather when all other predictors are zero. temp_max: Keeping other factors equal, a one-unit increase in temp_max is correlated with a substantial rise in the log-odds of the weather being sunny (1.074). temp_min: Keeping other factors equal, a one-unit rise in temp_min is correlated with a substantial drop in the log-odds of the weather being sunny (-0.580). wind: Keeping all other factors equal, a one-unit increase in wind is correlated with a small increase in the log-odds of sunny weather (0.752). log_precipitation: Keeping all other variables equal, a one-unit increase in log_precipitation is correlated with a little increase in the log-odds of the weather being sunny (0.481).
this code evaluates performance of model on both training and testing data set.
# Assuming 'data_train' is your data frame
data_train$weather <- as.factor(data_train$weather)
data_test$weather <- as.factor(data_test$weather)
# Now fit the model
my_mod <- multinom_reg() %>%
set_engine("nnet", trace = FALSE)
fit <- fit(my_mod, data = data_train, weather ~ temp_max + temp_min + wind + log_precipitation)
# Evaluate the model on the training data (classification)
train_results <- predict(fit, data_train, type = "class")
# Evaluate the model on the test data
test_results <- predict(fit, data_test, type = "class")
# Print the training results
train_results## # A tibble: 1,168 × 1
## .pred_class
## <fct>
## 1 rain
## 2 sun
## 3 sun
## 4 rain
## 5 rain
## 6 sun
## 7 sun
## 8 rain
## 9 rain
## 10 rain
## # ℹ 1,158 more rowstest_results## # A tibble: 293 × 1
## .pred_class
## <fct>
## 1 rain
## 2 snow
## 3 snow
## 4 snow
## 5 snow
## 6 rain
## 7 rain
## 8 rain
## 9 sun
## 10 rain
## # ℹ 283 more rowsmy_wf <- workflow() %>%
add_model(my_mod) %>%
add_recipe(my_rec)
# Fit the workflow to the training data
fit <- fit(my_wf, data_train)
# Print the fit object
fit## ══ Workflow [trained] ══════════════════════════════════════════════════════════
## Preprocessor: Recipe
## Model: multinom_reg()
##
## ── Preprocessor ────────────────────────────────────────────────────────────────
## 0 Recipe Steps
##
## ── Model ───────────────────────────────────────────────────────────────────────
## Call:
## nnet::multinom(formula = ..y ~ ., data = data, trace = ~FALSE)
##
## Coefficients:
## (Intercept) temp_max temp_min wind log_precipitation
## fog 3.035856 0.01800714 0.2543537 0.1913660 2.7186955
## rain 13.163772 -0.58936404 0.5849886 0.9962782 15.5386525
## snow -2.715052 -3.86803531 -3.3271336 1.5239656 19.1788878
## sun 3.045500 1.07381872 -0.5795547 0.7517946 0.4814626
##
## Residual Deviance: 1054.584
## AIC: 1094.584readr::spec## function (x)
## {
## stopifnot(inherits(x, "tbl_df"))
## attr(x, "spec")
## }
## <bytecode: 0x0000025f7c1d9060>
## <environment: namespace:readr>install.packages("parsnip")## Warning: package 'parsnip' is in use and will not be installedinstall.packages("workflows")## Warning: package 'workflows' is in use and will not be installedinstall.packages("yardstick")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'yardstick' successfully unpacked and MD5 sums checked## Warning: cannot remove prior installation of package 'yardstick'## Warning in file.copy(savedcopy, lib, recursive = TRUE): problem copying
## C:\Users\sivak\AppData\Local\R\win-library\4.3\00LOCK\yardstick\libs\x64\yardstick.dll
## to
## C:\Users\sivak\AppData\Local\R\win-library\4.3\yardstick\libs\x64\yardstick.dll:
## Permission denied## Warning: restored 'yardstick'##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packagesinstall.packages("tidymodels")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'tidymodels' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packages# Load the installed packages
library(parsnip)
library(workflows)
library(yardstick)## Warning: package 'yardstick' was built under R version 4.3.1##
## Attaching package: 'yardstick'## The following object is masked from 'package:readr':
##
## speclibrary(tidymodels)## Warning: package 'tidymodels' was built under R version 4.3.1## ── Attaching packages ────────────────────────────────────── tidymodels 1.1.1 ──## ✔ broom 1.0.5 ✔ tibble 3.2.1
## ✔ dials 1.2.0 ✔ tidyr 1.3.0
## ✔ infer 1.0.5 ✔ tune 1.1.2
## ✔ modeldata 1.2.0 ✔ workflowsets 1.0.1
## ✔ purrr 1.0.2## Warning: package 'dials' was built under R version 4.3.1## Warning: package 'modeldata' was built under R version 4.3.1## Warning: package 'purrr' was built under R version 4.3.1## Warning: package 'tune' was built under R version 4.3.1## Warning: package 'workflowsets' was built under R version 4.3.1## ── Conflicts ───────────────────────────────────────── tidymodels_conflicts() ──
## ✖ purrr::discard() masks scales::discard()
## ✖ tidyr::extract() masks magrittr::extract()
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ✖ purrr::set_names() masks magrittr::set_names()
## ✖ yardstick::spec() masks readr::spec()
## ✖ recipes::step() masks stats::step()
## • Use suppressPackageStartupMessages() to eliminate package startup messages# Define my_mod, my_rec, and train_test_split
# Perform the last_fit operation
final_fit <- last_fit(my_mod, my_rec, split = train_test_split)
final_fit## # Resampling results
## # Manual resampling
## # A tibble: 1 × 6
## splits id .metrics .notes .predictions .workflow
## <list> <chr> <list> <list> <list> <list>
## 1 <split [1168/293]> train/test split <tibble> <tibble> <tibble> <workflow>options(scipen = 999)
final_fit %>%
collect_predictions()## # A tibble: 293 × 10
## id .pred_drizzle .pred_fog .pred_rain .pred_snow .pred_sun .row
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <int>
## 1 train/test spl… 2.47e-16 2.15e-13 0.968 0.0317 1.59e-14 2
## 2 train/test spl… 1.76e-15 7.87e-13 0.380 0.620 7.97e-14 14
## 3 train/test spl… 2.92e-17 2.06e-14 0.103 0.897 1.53e-15 17
## 4 train/test spl… 2.62e-20 2.65e-17 0.00169 0.998 9.49e-19 18
## 5 train/test spl… 1.85e-18 9.95e-16 0.00453 0.995 9.32e-18 19
## 6 train/test spl… 4.28e-15 2.06e-12 0.876 0.124 2.77e-13 26
## 7 train/test spl… 5.71e-18 9.71e-15 0.949 0.0515 3.02e-16 29
## 8 train/test spl… 1.11e-13 3.46e-11 0.999 0.00120 1.68e-12 39
## 9 train/test spl… 1.51e- 1 1.74e- 1 0.0517 0.00000425 6.23e- 1 46
## 10 train/test spl… 8.77e-14 2.47e-11 0.954 0.0456 1.58e-12 51
## # ℹ 283 more rows
## # ℹ 3 more variables: .pred_class <fct>, weather <fct>, .config <chr>from the above results, it has given what is the predicted class and what is the actual weather condition. it gave different probabilities for different classes for each row.
for example in the first row predicted _drizzle is very less. model has predicted it as rain with 96% and the actual condition also rain. same way for all rows.
# Install the required packages if you haven't already
install.packages("parsnip")## Warning: package 'parsnip' is in use and will not be installedinstall.packages("workflows")## Warning: package 'workflows' is in use and will not be installedinstall.packages("yardstick")## Warning: package 'yardstick' is in use and will not be installedinstall.packages("tidymodels")## Warning: package 'tidymodels' is in use and will not be installed# Load the installed packages
library(parsnip)
library(workflows)
library(yardstick)
library(tidymodels)
# Assuming you have already defined your `my_wf` workflow and `data_train` and `data_test` datasets
# Fit the model using the training data
trained_model <- last_fit(fit, split = train_test_split)
# Evaluate the model on the testing data
testing_results <- collect_metrics(trained_model, data = train_test_split)
# Print the evaluation metrics (e.g., accuracy, ROC AUC)
testing_results## # A tibble: 2 × 4
## .metric .estimator .estimate .config
## <chr> <chr> <dbl> <chr>
## 1 accuracy multiclass 0.850 Preprocessor1_Model1
## 2 roc_auc hand_till 0.858 Preprocessor1_Model1this model has given accuracy of 84.98 % and roc_auc as 0.857. means the total corrected predictions are84.98%.
# Extract predictions from the trained model
predictions <- collect_predictions(trained_model, data = data_test)
# Calculate accuracy
accuracy_result <- accuracy(predictions, truth = weather, estimate = .pred_class)
# Print the accuracy result
accuracy_result## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 accuracy multiclass 0.850# Extract predictions from the trained model
predictions <- collect_predictions(trained_model, data = data_test)
# Calculate precision, recall, and F1 score
precision_result <- precision(predictions, truth = weather, estimate = .pred_class)## Warning: While computing multiclass `precision()`, some levels had no predicted events (i.e. `true_positive + false_positive = 0`).
## Precision is undefined in this case, and those levels will be removed from the averaged result.
## Note that the following number of true events actually occured for each problematic event level:
## 'drizzle': 9
## 'fog': 21recall_result <- recall(predictions, truth = weather, estimate = .pred_class)
f1_result <- f_meas(predictions, truth = weather, estimate = .pred_class)## Warning: While computing multiclass `precision()`, some levels had no predicted events (i.e. `true_positive + false_positive = 0`).
## Precision is undefined in this case, and those levels will be removed from the averaged result.
## Note that the following number of true events actually occured for each problematic event level:
## 'drizzle': 9
## 'fog': 21# Print the precision, recall, and F1 score results
precision_result## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 precision macro 0.905recall_result## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 recall macro 0.509f1_result## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 f_meas macro 0.856from this precision is 0.9053 and recall is 0.509 and f1 score is 0.855.
Naive-bias model
I want to try with naive bias model and i will check for accuracy, precision, recall,f1 score.
install.packages("naivebayes")## Installing package into 'C:/Users/sivak/AppData/Local/R/win-library/4.3'
## (as 'lib' is unspecified)## package 'naivebayes' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\sivak\AppData\Local\Temp\Rtmp44Faus\downloaded_packageslibrary(naivebayes)## Warning: package 'naivebayes' was built under R version 4.3.1## naivebayes 0.9.7 loaded# Load the required data and libraries
library(naivebayes)
# Assuming 'data_train' is your training data and 'data_test' is your test data
# 'weather' is the target variable, and 'temp_max', 'temp_min', 'wind', and 'log_precipitation' are your predictor variables
# Create and fit the Naive Bayes model
nb_model <- naive_bayes(weather ~ temp_max + temp_min + wind + log_precipitation, data = data_train)
# Predict on the test data
nb_predictions <- predict(nb_model, newdata = data_test)## Warning: predict.naive_bayes(): more features in the newdata are provided as
## there are probability tables in the object. Calculation is performed based on
## features to be found in the tables.# Print the predicted classes
nb_predictions## [1] rain snow snow snow snow rain rain rain sun
## [10] rain sun rain rain rain rain snow rain sun
## [19] rain rain rain rain rain rain rain rain sun
## [28] sun sun sun rain rain rain sun rain sun
## [37] sun sun sun sun rain rain sun sun sun
## [46] sun sun sun sun sun sun sun sun sun
## [55] sun sun sun sun sun rain rain sun rain
## [64] rain sun rain rain rain rain sun sun rain
## [73] rain rain rain rain snow snow rain rain rain
## [82] sun rain rain sun sun rain rain rain sun
## [91] rain rain sun sun rain rain sun sun sun
## [100] sun sun rain sun sun rain rain rain sun
## [109] sun rain sun sun sun sun sun sun rain
## [118] sun sun sun sun sun sun sun sun rain
## [127] rain sun sun rain sun rain rain rain rain
## [136] rain rain sun rain sun sun sun sun rain
## [145] rain snow rain sun rain sun sun sun rain
## [154] sun rain rain rain rain rain rain sun rain
## [163] sun sun sun sun sun rain rain sun rain
## [172] sun sun sun sun sun sun sun sun rain
## [181] rain rain sun sun sun rain sun sun sun
## [190] sun rain sun sun sun sun sun sun sun
## [199] rain rain sun sun rain rain rain sun sun
## [208] sun rain rain rain rain sun sun rain sun
## [217] rain rain rain rain sun rain rain sun rain
## [226] sun sun sun sun rain rain rain sun rain
## [235] sun sun sun rain sun sun sun sun sun
## [244] sun sun sun sun sun sun sun sun sun
## [253] sun sun sun sun sun sun rain sun sun
## [262] rain rain rain rain rain sun sun sun sun
## [271] sun rain sun sun sun rain rain rain rain
## [280] rain rain sun sun sun drizzle rain rain rain
## [289] rain rain snow rain sun
## Levels: drizzle fog rain snow sun# Convert predicted classes to character type
nb_predictions <- as.character(nb_predictions)
# Convert true class labels in test data to character type
true_labels <- as.character(data_test$weather)
# Calculate accuracy
accuracy <- sum(nb_predictions == true_labels) / length(nb_predictions)
# Create a confusion matrix
confusion_matrix <- table(true_labels, nb_predictions)
# Calculate precision, recall, and F1 score
precision <- diag(confusion_matrix) / rowSums(confusion_matrix)## Warning in diag(confusion_matrix)/rowSums(confusion_matrix): longer object
## length is not a multiple of shorter object lengthrecall <- diag(confusion_matrix) / colSums(confusion_matrix)
f1_score <- 2 * (precision * recall) / (precision + recall)## Warning in precision * recall: longer object length is not a multiple of
## shorter object length## Warning in precision + recall: longer object length is not a multiple of
## shorter object length# Print the metrics
cat("Accuracy:", accuracy, "\n")## Accuracy: 0.8395904cat("Precision:", precision, "\n")## Precision: 0 0 0.02158273 0 0cat("Recall:", recall, "\n")## Recall: 0 0 0.3333333 0cat("F1 Score:", f1_score, "\n")## F1 Score: NaN NaN 0.04054054 NaN NaNfrom the above i got results as accuracy as 0.839 and precision as 0.0215 and recall as 0.333.
so when we compare with naive bias and multi linear logistic regression model. Multi linear logistic regression is best.