Comparison of the Effectiveness of Machine Learning Models and Statistical Models in Predicting Financial Risk

The purpose of this study was to compare the efficiency of machine learning models (32 models) and statistical models (14 models) in predicting the financial risk of listed 145 companies in Tehran Stock Exchange during the period 2010 to 2020 and selecting the best model using advanced optimization techniques. Findings of the research using the test of comparing the accuracy of prediction coefficients, indicates that with 99 percent confidence, the prediction accuracy of machine learning models is higher than statistical models. Also, the best machine learning model after optimization was the evolutionary support vector machine model with 99.86 percent prediction accuracy and the value of the area under the curve was 0.998. In addition, accrual financial ratios with 99.45 percent predictive accuracy and operating financial ratios with 98.62 percent predictive accuracy were able to perform better than other financial ratios in using the evolutionary support vector to predict financial risk. on the other side, the projected financial risk varied according to different industries. Therefore, it was found that machine learning models can be used as an important tool in predicting corporate financial risk due to the lack of limitations that statistical models face.