Algorithmic Issues and Fraud Risks of Intelligent Investment Advice on FinTech Platforms: a Study Combining Investor Behavior

Abstract

The article constructs a fusion intelligent investment advisor fraud identification and asset recommendation model, which captures user behavior sequence features based on LSTM, combines with XGBoost to process structured transaction data, and achieves multi-dimensional risk judgment and return optimization through integrated learning. Based on the traditional rule engine and single model, the recognition accuracy and configuration precision are improved. The experimental results show that the F1-score of the fusion model reaches 0.848 and the AUC is improved to 0.902 in the fraud recognition task; in the asset recommendation task, the return of the aggressive investor is improved to 17.91%, which is better than the performance of the other models, which verifies the effectiveness and adaptability of the method.