In Python: $1.$ Set predictors and outcome variables, and partition the data into $60\%$ training, $40\%$ validation

#$1$

#Set predictors and outcome

predictors $=[\text{'}$Cleanliness$\text{'},$ 'Nurse communication, 'Doctor Communication', 'Discharge information', 'Met$\_$Color', 'Automatic', $\text{'}$CC$\text{'},$ 'Doors', 'Tax', 'Weight'$]$

outcome $=$ 'Rating'

# $2.$ Train a tree for classification

# $3.$ Run random forest and boosted tree models for classification

# $4.$ Which rf variable is the most important? $($e$.$g$.,$ importances $=$ rf$\_$model.named$\_$steps$[\text{'}$classifier$\text{'}].$feature$\_$importances$\_)$

# $5.$ Assess the performance of all models using Sklearn classification$\_$report. Which model is better? See an example below

#of how to set up one of the models

# classTree $=$ DecisionTreeClassifier$($random$\_$state$=0,$ max$\_$depth $=5,$ min$\_$samples$\_$split $=50)$

# tree$\_$model $=$ Pipeline$($steps$=[$

# $(\text{'}$preprocessor$\text{'},$ preprocessor$),$

# $(\text{'}$classifier$\text{'},$ classTree$)$

# $])$

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

# pred$\_$y $=$ tree$\_$model.predict$($valid$\_$X$)$

# print$(\text{'}--------------$DecisionTreeClassifier Classification Report$--------------------\text{'})$

# print$($classification$\_$report$($valid$\_$y$,$ pred$\_$y$))$

# $6.$ Visualize the tree. See an example below

# fitted$\_$tree $=$ tree$\_$model.named$\_$steps$[\text{'}$classifier$\text{'}]$

# class$\_$names $=[$str$($cls$)$ for cls in fitted$\_$tree.classes$\_]$

# transformed$\_$feature$\_$names $=$ tree$\_$model.named$\_$steps$[\text{'}$preprocessor$\text{'}].$get$\_$feature$\_$names$\_$out$()$

# dot$\_$data $=$ export$\_$graphviz$($fitted$\_$tree, out$\_$file$=$None,

# feature$\_$names$=$transformed$\_$feature$\_$names,

# class$\_$names$=$class$\_$names,

# filled$=$True, rounded$=$True,

# special$\_$characters$=$True$)$

# graph $=$ graphviz.Source$($dot$\_$data$)$

# graph.render$("$Decision$\_$Tree", format$="$png$"$

# graph.view$()$