Breaking News: Grepper is joining You.com. Read the official announcement!

ML

Rui Lin answered on January 9, 2024 Popularity 1/10 Helpfulness 1/10

answer ML

ML

Comment

def calculate_cost(left_subset, right_subset, classes):
    total_data_points = len(left_subset) + len(right_subset)

    gini_left = gini_right = 0

    if len(left_subset) > 0:
        sum_of_prob = 0
        for c in classes:
            # find the num of data points which are in class c
            num_in_c = len([l for l in left_subset if l[1] == c])
            # now we can get the probability of class c in this subset
            prob = num_in_c / len(left_subset)
            # sum this value squared as per the equation
            sum_of_prob += prob**2
        gini_left = 1 - sum_of_prob

    # we do the same for the right subset
    if len(right_subset) > 0:
        sum_of_prob = 0
        for c in classes:
            num_in_c = len([l for l in right_subset if l[1] == c])
            prob = num_in_c / len(right_subset)
            sum_of_prob += prob**2
        gini_right = 1 - sum_of_prob

    # we calculate the cost
    cost = (len(left_subset) / total_data_points) * gini_left + (
        len(right_subset) / total_data_points
    ) * gini_right

    return cost


def decision_tree_best_split(X, y):
    best_feature_val = best_feature_idx = best_cost_val = float("inf")
    best_data_point = None
    classes = range(len(set(y)))  # get the number of classes

    # Iterate over each possible feature value
    for feature in range(len(X[0])):
        for data_point in range(len(X)):
            feature_threshold_val = X[data_point][feature]
            # based on our threshold value, we determine
            # which data points go into the left or right subset
            left_subset = []
            right_subset = []
            for instance, label in zip(X, y):
                if instance[feature] < feature_threshold_val:
                    left_subset.append((instance, label))
                else:
                    right_subset.append((instance, label))
            # calculate the cost
            curr_cost = calculate_cost(
                left_subset, right_subset, classes
            )  # TALK ABOUT REMOVING CLASSES

            # update our best feature if the current cost
            # is less than our best so far
            if curr_cost < best_cost_val:
                best_feature_idx = feature
                best_feature_val = feature_threshold_val
                best_cost_val = curr_cost
                best_data_point = data_point

    return (best_feature_val, best_feature_idx, best_data_point)

Popularity 1/10 Helpfulness 1/10 Language python

Source: www.machinelearninginterviews.com

Tags: ml python

Link to this answer
Share Copy Link

Contributed on Jan 09 2024

Rui Lin

0 Answers Avg Quality 2/10

ML

Contents

More Related Answers

ML

Grepper

Documentation

Social

Legal

Contact

Oops, You will need to install Grepper and log-in to perform this action.