Financial fraud detection based on random forest and XGBoost algorithm

In the digital age, emerging business forms such as Internet finance, mobile payment, and digital currency have not only promoted the rapid development of the financial industry but also provided new ground and more concealed means for fraudulent acts. The article selects the Random Forest and XGBoost algorithms to detect financial fraud. It analyzes the financial log data provided by a multinational financial service provider, screens its feature engineering from five perspectives, and draws conclusions such as that fraudulent transactions mostly occur in two types of transactions, namely transfer and cash withdrawal, the time when frauds happen is relatively concentrated at certain special time points, and fraudulent transactions are more likely to appear among non-merchant users. Based on this, the article also puts forward several targeted countermeasures and suggestions for preventing fraudulent acts.