摘要
Credit card fraud detection targets to find out credit card fraud by analyzing the features of transaction data, which can prevent huge financial losses and reduce the pressure on police. There are three challenges in credit card fraud detection, i.e., unbalanced data, excessive invisible features, and the choice of suitable threshold to make the models more reasonable. To overcome the above challenges, this paper uses a modified Logistic Regression (LR) model to identify credit card frauds. Firstly, we balance the data by adopting the SMOTE method. We also figure out features importance and remove unimportant features. Secondly, we train the Logistic Regression model, and find out the best parameters for the model by Grid CV Search and compare their performance. Finally, we select the best threshold of the model by comparing confusion matrixes. To verify the effectiveness of LR, we compare LR with Support Vector Machines (SVM), K-Nearest Neighbour (KNN) and Decision Tree Classifier (DT) with the model score and Area Under ROC Curve (AUC). The experimental results show that LR achieves a 0.94 accuracy score, while KNN, SVM and DT achieve 0.93, 0.93 and 0.91 with default parameters, respectively. LR gets the highest classifier score 0.9824 at the AUC score, which demonstrates LRs effectiveness in credit card fraud prediction.
学科与关键词
- The Optimized Anomaly Detection Models Based on an Approach of Dealing with Imbalanced Dataset for Credit Card Fraud DetectionZhang, Yan-Feng; Lu, Hong-Liang; Lin, Hong-Fan; Qiao, Xue-Chen; Zheng, Hao*Mobile Information Systems, 2022, 2022: 8027903.被引 3
- Credit card fraud detection in the era of disruptive technologies: A systematic reviewCherif, Asma*; Badhib, Arwa; Ammar, Heyfa; Alshehri, Suhair; Kalkatawi, Manal; Imine, AbdessamadJournal of King Saud University-Computer and Information Sciences, 2023, 35(1): 145-174.被引 31
- Credit Risk Models for Financial Fraud Detection: A New Outlier Feature Analysis Method of XGBoost With SMOTEXia, Huosong; An, Wuyue; Zhang, Zuopeng*JOURNAL OF DATABASE MANAGEMENT, 2023, 34(1).被引 3
