Abstract

In this research, the fraud detection model was trained using Matlab R2022a, with the dataset being randomly split into training and testing sets at a 7:3 ratio. Specifically, 70% of the data was allocated for training purposes, while the remaining 30% was used for testing. Upon incorporating the XGBoost model, the confusion matrix analysis indicated that a total of 6,326,133 instances were accurately predicted, with only 36,487 instances misclassified in the test set. This translates to an outstanding model performance, achieving a prediction accuracy of 99%. This demonstrates that the model exhibits robust performance and reliability in detecting fraudulent activities. Through this study, we have not only enhanced our capacity to identify diverse types of fraud but also provided valuable insights for future optimization of fraud detection techniques. The findings hold significant importance for safeguarding personal and organizational financial security and will contribute positively to the development of a more secure and reliable financial and network ecosystem.