Fraud Detection in Online Transaction using Hybrid CNN and XGBOOST

Online transactions have become an essential part of our daily life; this makes such transactions more probable to result in fraud. The users card details have been abused by fraudsters and online transactions are made through them. The scammers are always looking for new tools and ways to carry out their fraudulent activities. Machine learning algorithms have the capacity to develop and identify fraud patterns that were not previously observed. In this study, datasets of financial transactions are applied in various machine learning algorithms to solve the problem of transaction fraud. These techniques are based on the hybridization form of Convolution Neural Network and Support Vector Machine (CNN-SVM) and Extreme Gradient Boosting algorithm (XGBOOST). The SVM classifier was used to replace the CNN architecture model’s final last output layer in order to create the hybrid CNN-SVM model architecture. An enhanced variation of the gradient-boosting strategy is called XGBOOST. It enhances the system’s performance by employing ensemble approaches. To address an unbalanced class distribution, the current classification models are modified using ensemble approaches. The XGBoost model and the hybrid model are contrasted. We demonstrate that our suggested model can identify fraudulent transactions with a high degree of accuracy and a comparatively low quantity of false positives.