Intelligent Anomaly Detection in Financial Transactions Using Machine Learning Paradigms

Abstract

The main objective of the project is to detect fraudulent activity in financial data using machine learning techniques. In the banking industry, where identifying and stopping fraudulent transactions is crucial, this is a serious issue. The research presents class weight-tuning hyperparameters to enhance fraud detection. By improving the model's ability to distinguish between authentic and fraudulent transactions, these parameters raise the fraud detection system's accuracy. Three well-known machine learning algorithms—XGBoost, LightGBM, and CatBoost—are explicitly used in the study. The goal of combining the benefits of each algorithm is to improve the overall efficacy of the fraud detection technique.Hyperparameters are optimized in the study using deep learning techniques. This link improves the fraud detection system's efficacy and flexibility, increasing its capacity to recognize changing fraud strategies. The initiative uses real-world data to do comprehensive assessments. These studies demonstrate that the combination of LightGBM and XGBoost works better than current techniques across a range of parameters. This suggests that the suggested method outperforms alternative methods in terms of identifying fraudulent activity. A Stacking Classifier is one feature that combines the predictions of the RandomForest and LightGBM classifiers with certain settings. By combining the advantages of many models, this ensemble method improves prediction accuracy by using a GradientBoostingClassifier as the final estimator.