marwaomar / ml_assignment_3_decision-trees-

Task 3 (Decision Trees )

1. Loading the data and did some preprocessing on it such as assuming categories for continous data and filling the missing data or adjusting the outliers

Then the data looks finally like :

2.Training and test split to 90/10

3.Started defining function:

Gini_impurity :

takes lable and label index, it counts how many unique values in the target and each unique value counts

information_gain: Calculates info gain from the following formula:

SplitDataSet:

Splitting data according to a value

find_best_split:

for each feature it takes the values and split it then calculate varience to get best feature and best value

Count:

to get unique values and its count

Leaf ,Decision_Note :

Two classes to help building and printing the tree which i got from :

Github

build_tree:

Using all the previous functions to finaly build a tree and use a recursive method to repeat for each branch

Predict:

first it checks if we are at the leaf node then it keeps asking where is the new test data recursively .

Accuracy:

It measure the matches between predicted values and true values over the length of the data.

Finally i tried sklearn code to compare both codes.

Results :