#Train Test Split

from sklearn.model$\_$selection import train$\_$test$\_$split

X$\_$train, X$\_$test, y$\_$train, y$\_$test $=$ train$\_$test$\_$split$($X$\_$new, y$,$ test$\_$size$=0\mathrm{.}3,$ random$\_$state$=123)$

#$30\%$ of the given data is used as testing data, the remaining $70\%$ is training data. This selection was made randomly.

#$\%\%$

from sklearn.ensemble import RandomForestClassifier

rf$\_$default $=$ RandomForestClassifier$($random$\_$state$=123)$

rf$\_$default.fit$($X$\_$train, y$\_$train$)$

y$\_$predict$\_$rf $=$ rf$\_$default.predict$($X$\_$test$)$

#$\%\%$

def evaluate$\_$model$($y$\_$predict, y$\_$test$)$:

# Evaluate the performance of model using the test data.

# Use accuracy score, precision, recall and confusion matrix as performance metrics.

confusion$\_$matrix$\_=$ metrics.confusion$\_$matrix$($y$\_$test, y$\_$predict$)$

sns$.$heatmap$($confusion$\_$matrix$\_,$ annot$=$True, fmt$="$d$")$

print$("$Accuracy: $","\{$:$\mathrm{.}2$f$\}".$format$($metrics$.$accuracy$\_$score$($y$\_$test, y$\_$predict$)),$

$"$

Precision: $","\{$:$\mathrm{.}2$f$\}".$format$($metrics$.$precision$\_$score$($y$\_$test, y$\_$predict$)),$

$"$

Recall: $","\{$:$\mathrm{.}2$f$\}".$format$($metrics$.$recall$\_$score$($y$\_$test, y$\_$predict$)),$

$"$

Confusion Matrix: $")$

#$\%\%$

evaluate$\_$model$($y$\_$predict$\_$rf$,$ y$\_$test$)$

#We obtain highest accuracy level, precision and recall. However, we can use grid search cross validation to check our models performance again.

#Accoding to Breiman $(2001),$ who proposed Random Forest, max$\_$features and n$\_$estimators ar most important parameters of Random Forest. We can try to optimize them.

#In addition to this $,$ we may try to balance the class weights to overcome imbalance data problem.

#$\%\%$

params $=\{$

'max$\_$features': $["$auto$",$ "sqrt", "log$2"],$

$\text{'}$n$\_$estimators': $[300,500,700,1000]$

$\}$

#$\%\%$

rf$\_$default $=$ RandomForestClassifier$($class$\_$weight$=$ "balanced$\_$subsample", random$\_$state$=123)$

stratified$\_$kfold $=$ StratifiedKFold$($n$\_$splits$=10,$ shuffle $=$ True, random$\_$state$=123)$

grid$\_$search $=$ GridSearchCV$($rf$\_$default, params, n$\_$jobs$=-1,$ cv$=$stratified$\_$kfold, verbose$=2)$

grid$\_$search$\_$results $=$ grid$\_$search.fit$($X$\_$new, y$.$values.ravel$())$

#$\%\%$

target $=$ 'class'

X $=$ mushroom$\_$data.drop$($columns$=[$target$])$

y $=$ mushroom$\_$data$[$target$]$

print$($f$\text{'}$Y shape $=\{$y$.$shape$\}\text{'})$

print$($f$\text{'}$X shape $=\{$X$.$shape$\}\text{'})$

#$\%\%$

from sklearn.model$\_$selection import train$\_$test$\_$split

X$\_$train, X$\_$test, y$\_$train, y$\_$test $=$ train$\_$test$\_$split$($X$,$ y$,$ test$\_$size$=0\mathrm{.}2,$ random$\_$state$=42)$

print$($f$\text{'}$shape of X Train $=\{$X$\_$train.shape$\}\text{'})$

print$($f$\text{'}$shape of X Test $=\{$X$\_$test.shape$\}\text{'})$

print$($f$\text{'}$shape of Y Train $=\{$y$\_$train.shape$\}\text{'})$

print$($f$\text{'}$shape of Y Test $=\{$y$\_$test.shape$\}\text{'})$

#$\%\%$

acc$\_$baseline $=$ y$\_$train.value$\_$counts$($normalize$=$ True$).$max$()$

print$($f$\text{'}$Accuracy of baseline $=\{$acc$\_$baseline$\}\text{'})$

#$\%\%$

from sklearn.preprocessing import OrdinalEncoder

from sklearn.pipeline import make$\_$pipeline

from sklearn.ensemble import RandomForestClassifier

clf $=$ make$\_$pipeline$($OrdinalEncoder$(),$

RandomForestClassifier$($random$\_$state$=42))$

params $=\{$

'randomforestclassifier$\_\_$n$\_$estimators': range$(25,100,25),$

'randomforestclassifier$\_\_$max$\_$depth': range$(10,70,10)$

$\}$

params

#$\%\%$

# summarize results

print$("$Best: $\%$f using $\%$s$"\%($grid$\_$search$\_$results.best$\_$score$\_,$ grid$\_$search$\_$results.best$\_$params$\_))$

#$\%\%$

from sklearn.model$\_$selection import GridSearchCV

model $=$ GridSearchCV$($

clf$,$

param$\_$grid$=$ params,

cv$=5,$

n$\_$jobs$=-1,$

verbose $=1$

$)$

model

#$\%\%$

model.fit$($X$\_$train $,$ y$\_$train$)$

#$\%\%$

cv$\_$results $=$ pd$.$DataFrame$($model$.$cv$\_$results$\_)$

cv$\_$results.sort$\_$values$($by$=$ 'rank$\_$test$\_$score'$)$

#$\%\%$

#evaluate$\_$model$($y$\_$pred, y$)$

cv$\_$results.sort$\_$values$($by$=$'rank$\_$test$\_$score'$)$

rf$\_$model $=$ RandomForestClassifier$($class$\_$weight$=$"balanced$\_$subsample", max$\_$features$=$'auto', n$\_$estimators$=300,$

random$\_$state$=123)$

rf$\_$default.fit$($X$,$ y$)$

y$\_$pred $=$ rf$\_$default.predict$($X$)$

#$\%\%$

features $=$ X$\_$test.columns

importances $=$ model.best$\_$estimator$\_.$named$\_$steps$[\text{'}$randomforestclassifier$\text{'}].$feature$\_$importances$\_$

feat$\_$imp $=$ pd$.$Series$($importances $,$ index $=$ features$).$sort$\_$values$()$

feat$\_$imp.tail$().$plot$($kind$=$ 'barh'$)$

plt$.$xlabel$("$Gini Importance"$)$

plt$.$ylabel$("$Feature$")$

plt$.$title$("$Feature Importance"$)$; THE PYTHON CODE GIVEN ABOVE IS RELATED TO RANDOM FOREST CLASSIFICATION IN THE DATA SCIENCE COURSE.

PLEASE INTERPRET THIS CODE AND PREPARE A REPORT and presentation ACCORDING TO THE SUBJECTS AND CODES.