FRAUD DETECTION IN FINANCIAL TRANSACTIONS USING TREE-BASED MACHINE LEARNING MODEL

• Main Author: Pratama Putra, Arya
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77597
• Institution: Institut Teknologi Bandung

In recent decades, technological developments have led to the rise of e-commerce and transactions over the internet. The popularity of online transactions worldwide has attracted criminals to commit financial fraud in online transactions. This shows the importance of fraud detection in online transactions. The purpose of this Final Project is to apply several tree-based machine learning models to detect financial fraud in online transactions using the dataset provided by Vesta in the Kaggle competition organized by the IEEE Computation Intelligence Society (CIS), and then compare the performance of these models. This Final Project implements three models, namely Classification and Regression Trees (CART), Random forest, and Extreme Gradient Boosting (XGBoost). Since the class imbalance was extremely high, resampling method was applied and then compared with the performance of the model without resampling. It was observed that the overall performance of the models using resampling datasets was worse than without resampling, except for the random forest model. Overall, the XGBoost model showed the best performance with an AUC score of 0.92, followed by Random Forest with an AUC score of 0.90, while CART showed the poorest performance with an AUC score of 0.85.