Research on Financial Risk Prediction Model Based on News Text and LS-SVM

In order to achieve accurate and effective prediction of corporate financial risks, a financial risk prediction model based on news text and LS-SVM is proposed. The paper clarifies the targeted data standards for financial risk data collection by analyzing the financial risk prediction index system; combining news texts and using web crawling technology to extract unstructured enterprise financial risk data from financial websites; extracting news text topic features based on potential Dirichlet allocation, and achieving comprehensive financial risk feature fusion based on news text; based on the least squares support vector machine, a financial risk prediction model is constructed by taking the comprehensive financial risk features of the fused news text as input. The experimental results show that the highest accuracy of risk prediction generated by the design model is [Formula: see text]%, the average accuracy of full sample prediction is [Formula: see text]%, the minimum prediction time is [Formula: see text][Formula: see text]s, and the time efficiency reaches the highest value of [Formula: see text]%. This indicates that the use of the design model can effectively capture the relationship between financial indicators and financial risks, ensure the accuracy of enterprise financial risk prediction, and have good scalability in large-scale data scenarios, with a relatively short overall risk prediction time.