Research Of Financial Enterprise Risk Prediction Based On Neural Network And Particle Swarm Optimization

Under the background of globalization of financial development,the international financial market has frequent business contacts,the financial environment is more complex,and has recently been affected by COVID-19.The financial pressure of China’s financial enterprises continues to grow and is faced with all kinds of risks.The risk problem of enterprises will not only cause interest losses to enterprise investors,creditors,but also bring hidden dangers to the normal operation of enterprises.Under this complex background and uncertain conditions,in order to ensure the healthy and orderly development of China’s financial enterprises and the good operation of the market economy,it is of great significance to analyze the relevant data of enterprises,build an effective financial enterprise risk prediction model,predict the enterprise risk status in advance,and then provide early warning information for enterprise managers and help enterprises take measures to deal with risks in advance.This paper studies the risk prediction of financial enterprises from the financial risk of financial enterprises,selects the financial data of the first three years of A-share listed financial enterprises from 2002 to 2021 as the risk sample,and then matches the samples of normal enterprises from 2002 to 2019 that have not been marked by the CSRC by the end of 2021 according to the same industry,the same period and the same asset scale.The risk prediction problem of financial enterprises is transformed into a binary classification problem.After that,this paper solves the problem of missing sample data,data imbalance and feature selection through KNN,smote and RF Gini algorithms,and selects 12 features with different degrees of risk of listed financial enterprises.Then,a hybrid algorithm model based on PSO(particle swarm optimization)-BP neural network is used to construct the risk prediction model of financial enterprises based on the samples of listed enterprises in T-3 years(the first three years since they are marked as St enterprises)and these 12 final characteristics.Through comparative experiments,it is found that the PSO-BP neural network algorithm model is better than the traditional BP network model in predicting the risk of financial enterprises.The overall accuracy of the hybrid model in the training set is 88.3%,and the overall accuracy in the test set is 86.2%.The prediction ability of the model is good.The experimental results show that the PSO-BP hybrid model can be better applied to the risk prediction of financial enterprises.