Company’s Financial Risk Prediction And Control Based On L1 Penalty Logit Model

In the context of economic globalization and China’s flourishing capital market, the financial status of listed companies has been drawing attention from various interest groups; and meanwhile the financial risk management has proved a crucial link in running a company. An appropriate model for the financial risk management can effectively help predict the occurrence and development of risks as well as their patterns of volatility, which is an instrumental in the oversight of the financial status on the part of top executives and stakeholders, and serves as reference to averting and regulating financial crises and increasing the level of risk management.This thesis constructs the L1 penalty Logit model through combining L1 penalty function and Logit model, and applies it to the company’s financial risk analysis. The model was able to judge the company’s financial risk status by bankruptcy probability prediction in the binary Logit analysis. Second, the model enables us to identify the key factors that affect the company’s financial risk through variable selection by L1 penalty term, which is helpful for risk control. The Monte Carlo simulations show that the L1 penalty Logit model can obtain higher accuracy of prediction than the classical support vector machine(SVM) with less explanatory variables or simpler model structure. In addition, we implement an empirical analysis on financial data of the ST and non-ST companies in Chinese stock market from 2005 to 2014. The empirical findings also show that the prediction accuracy of L1 penalty Logit model is higher compared to the SVM through confusion matrix and ROC curve. Moreover, those key factors selected by the L1 penalty Logit model provide a basis for controlling company’s financial risk.In this thesis, L1 penalty Logit model is applied to empirical analysis into corporate financial risks, from which it can conclude that L1 penalty Logit model help not only accurately predict a company’s financial risks and take better-targeted precautions; but identify critical factors from multitudinous ones and reduce the cost of risk oversight. These interesting results are expected to provide fundamental tools and reference for the company’s financial risk management and be conducive to enhancing the level thereof.