A$-4.$

$[10$ marks: $2\mathrm{.}5$ each$]$: Using the same dataset split in A$-3.$a

Page $2$ of $9$

ISE$-291$: Homework $04$

a$.$ Build a Random forest classifier for predicting the class label with $4$ trees. Fit the classifier

using the training set. Set criterion to entropy and random$\_$state to $62.$

b$.$ Draw the trees using sci$-$kit learn $($sklearn$)$

c$.$ Test the classifier on the testing data set, and print the confusion matrix and classification

metrics $($Accuracy$,$ sensitivity $($Recall$),$ Precision$)$ of the Random forest classifier.

d$.$ Repeat A$-4($a$-$c$)$ using a Random forest with $8$ trees instead of $4.$

A$-5.[10$ marks$]$: Calculate the Information Gain $($IG$)$ for the class variable $$Drug$$ given the feature

selected $$BP$$ as a root node.

A$-6.[10$ marks$]$: From the decision tree built in A$-3,$ write three classification rules using the

normalized values first then return it to the original values.

A$-7.[10$ marks$]$: Write an association rule for $"$ BP $->$ Cholestrol", which rule has the highest

accuracy? Write the corresponding support and accuracy.

A$-8.[10$ marks$]$: Repeat parts b$,$ c$,$ and d in A$-3$ using the Na$$ve Bayes GaussianNB classifier.

A$-9.$ Compare the performance of the Na$$ve Bayes against the built decision tree and random forest

classifiers using confusion matrix. Based on the comparison, which one is the best to use with

the given datat set?