Financial Risk Prediction Method Based on Texting Mining

Based on text mining, this paper prevents and controls financial risks by quantifying the logical relationship of financial risks, and at the same time deals with the problem of unreliable word frequency quantification of financial events. This paper proposes a quantitative analysis method of financial risk logical relationship based on BERT and mutual information combined with domain knowledge and quantifies the relationship in the general data set COPA and the financial domain data set. Based on BERT and mutual information, it can effectively solve the problem of unreliable word frequency quantification. The accuracy rate of this method in the quantification of financial risk logical relationship reaches 80.1%, which is 3.09~37.39% higher than the benchmark model. Only the corpus in the financial field is considered, and the effect on other non-financial corpus is pending inspection. This method can reveal the evolution path of financial risk events and improve the effect of quantifying the logical relationship of financial risk.