XGBoost-LLM Integrated Fraud Detection System

Abstract

In this study, the anomaly detection model was trained using Python with TensorFlow 2.8.0. The dataset was randomly partitioned into training and validation sets at a ratio of 8:2. That is, 80% of the data was dedicated to the training phase, and the remaining 20% was reserved for validation. After integrating the LightGBM model, the analysis of the confusion matrix showed that 8,215,462 cases were accurately predicted, and only 28,513 cases were misclassified in the validation set. This implies an excellent model performance, attaining a prediction accuracy of 99.6%. It clearly shows that the model has strong performance and stability in detecting abnormal activities. Through this research, we have not only improved our ability to recognize various types of anomalies but also offered useful guidance for the future improvement of anomaly detection methods. The results are of great significance for protecting individual and corporate information security and will make a positive contribution to the establishment of a more secure and trustworthy information and network environment.