Analysis of Tree-Based Machine Learning Techniques for Credit Card Fraud Detection
Jitender Tanwar, Shubham Singh, Akash Kumar, Mandeep Mittal, Leena Singh, Sudhanshu Tripathi · Apple Academic Press eBooks · 2023
A credit card is an easy and most beneficial target for hackers. On the other hand, it is a huge threat to the financial industry. Thousands of people suffer great losses from this problem every year. Although it is not an easy problem, machine learning (ML) has given us hope and played an important role in the fraud detection of online transactions. Online fraud detection is facing difficulties like having unbalanced data, not having confidential data, non-availability of analyzed data, etc. The processed and analyzed data can significantly improve the performance of ML algorithms. In this chapter, we analyzed and compared the performance of Random Forest, AdaBoost Classifier, XGBoost Classifier and finally ensembled them to get better results. 248 We have evaluated their performances based on their Matthews correlation coefficient (CC) score, AUC-ROC score, and their confusion matrix. The result analysis shows the order of importance of features for selected ML algorithms. The results indicate that the Random Forest has the best score in both the evaluation metrics which comes out to be 0.9981.