Addressing Digital Finance Challenges: The Application and Op-timization of Logistic Regression Models in Fraud Detection Sys-tems

The burst growth of electronic finance brings forward increasingly more sophisticated financial crimes and stricter requirements of anti-fraud technology. This paper presents the Logistic Regression based financial transaction fraud detection model which realizes a detection and identification process of such financial crimes based on the transaction data. The multiple features of transaction data are utilized, and logistic regression algorithm is adopted for model training to form a concise and effective financial fraud detection application.Through the analysis based on the experimental result, we conclude that the logistictic regression model can classify the minority class of fraudulent transaction effectively in the data imbalance situation, in addition, the model is with great interpretability and low computational overhead, but has some limitations in dealing with the complex nonlinear fraudulent behavior, has poor adaptability to the new fraudulent approach.We can also explore to introduce deep learning and ensemble learning approach in the future work to improve the accuracy of the model and dynamic adjustablity.