3N4N/adaboost

An assignment on Adaboost algorithm with Logistic Regression as weak learner

Logistic Regression and Adaboost for Classification

In this assignment of CSE472 course we implement adaboost algorithm with logistic regression as weak learner to predict the binary labels of given datasets.

Datasets Used

• https://www.kaggle.com/blastchar/telco-customer-churn
• https://archive.ics.uci.edu/ml/datasets/adult
• https://www.kaggle.com/mlg-ulb/creditcardfraud

The first two datasets are saved in this repo. The third one is too large to push to GitHub, so you'll need to download it yourself.

wget -O creditcard.zip https://www.kaggle.com/mlg-ulb/creditcardfraud/download
unzip creditcard.zip
mv creditcard.csv ./data/

Report of Logistic Regression

Telco dataset

Performance measure	Training	Test
Accuracy	80%	79%
True positive rate	53%	50%
True negative rate	89%	89%
Positive predictive value	65%	63%
False discovery rate	34%	36%
F1 score	58%	56%

Adult dataset

Performance measure	Training	Test
Accuracy	84%	84%
True positive rate	58%	58%
True negative rate	92%	92%
Positive predictive value	72%	71%
False discovery rate	27%	28%
F1 score	64%	64%

Credit Card Fraud dataset

Performance measure	Training	Test
Accuracy	42%	42%
True positive rate	94%	93%
True negative rate	41%	41%
Positive predictive value	03%	03%
False discovery rate	96%	96%
F1 score	07%	07%

Report of Adaboost

Telco dataset

Number of boosting rounds	Training	Test
5	79%	78%
10	81%	79%
15	80%	79%

Adult dataset

Number of boosting rounds	Training	Test
5	84%	84%
10	84%	84%
15	84%	84%

Credit Card Fraud dataset

Number of boosting rounds	Training	Test
5	78%	78%
10	98%	98%
15	97%	97%