Detection of Credit Card Fraud Using Resampling and Boosting Technique

Abstract

There are more online card transactions as a result of the development of technologies like financial technology and e-commerce applications. Fraud on credit cards has skyrock-eted, as a result affecting credit card companies, customers, retailers, and banks. Therefore, it is crucial to create systems that guarantee the confidentiality and accuracy of credit card transactions. Using Sparkov's imbalanced synthetic dataset, a Machine Learning (ML)-based remedy for fraud detection using credit cards is developed. Use of a synthetic dataset with named attributes distinguishes this proposed approach from the usual implementation of ML models using publicly available European Credit Card Fraud(CCF) Dataset in Kaggle. Most of the studies in this field are done with this dataset as a basis. To tackle the problem of class imbalance, the Synthetic Minority Oversampling Technique(SMOTE) named approach is used. The framework is then evaluated using Extra Tree (ET), Gradient Boost (GB), Decision Tree(DT) and Random Forest (RF). The above ML models were combined with the ensemble model known as AdaBoost to improve the quality of automated detection. Accuracy, Recall, Precision and F1-score are used to assess the models. To further validate the findings of this study, the suggested framework was applied to a synthetic dataset of credit card transactions which are highly skewed. The ML algorithms are paired with the AdaBoost algorithm to check its impact on the efficiency of the proposed method.