Machine Learning Algorithms in Financial Market Risk Prediction

Abstract

In the current complex international environment, whether it is data volume, data quality or computing power, the era of big data is different from the past. In addition to data problems, there are also algorithm problems that restrict the development of financial risk control, and the rapid development of machine learning just makes up for this deficiency. The purpose of this paper is to comprehensively compare the default risk early warning effects of machine learning algorithms and traditional statistical Logistic models, and to further select the most appropriate financial risk early warning model in the machine learning algorithm, so as to provide a scientific basis for online lending platforms to select default rate early warning models. method. This paper mainly studies the comparison of the prediction effects of the four models. In the test set, the early warning accuracy rate of the Logistic model is only 62.74%, while the early warning accuracy rate of the machine learning model is generally between 83% and 93%, and the accuracy rate of the integrated algorithm is generally around 90%. The lowest accuracy rate of the Stacking algorithm It also reached 85.8%, while the accuracy of the base classification algorithm was lower than 83%. The results show that, based on the specific scenario of loan risk assessment, the early warning accuracy of the machine learning algorithm is generally higher than that of the Logistic model, and it is more suitable for financial market risk prediction.