Enhancing Financial Fraud Detection: The Efficacy of Convolutional Neural Networks

Abstract

This study aims to explore and compare the application effects of various machine learning models in the task of financial fraud transaction detection, focusing on the analysis of the advantages of convolutional neural networks (CNN) in dealing with complex transaction data. In the experiment, we compared the performance of six models: logistic regression, support vector machine, random forest, K nearest neighbor, gradient boosting tree and CNN. The results show that CNN can more effectively identify abnormal patterns in financial transaction data with its powerful feature extraction ability, showing the highest accuracy, recall, precision and F1 score. However, due to the high-dimensional and large-scale characteristics of financial transaction data, the computational cost of CNN is high, so it is necessary to balance model performance and resource consumption in practical applications. This study not only verifies the potential of CNN in financial fraud detection, but also provides a research direction for future financial security applications based on deep learning. In the future, multi-source data can be further combined with emerging adaptive learning methods to improve the detection accuracy and robustness of the model.