With the continuous development of the financial industry, the prediction of the stock market is becoming more and more important in this industry, and more and more scholars are investing in the study of stock forecasting. The stock market risk premium has always been a classic problem in the financial industry. The reason is that common prediction models have unstable parameters and other problems that will affect prediction. Based on the current situation of stock prediction research in machine learning, this paper will analyze the operation principles of different prediction models, and introduce the characteristics, benefits and drawbacks of every type. Finally, The stock price is predicted by the Support Vector Machine model, the yield is predicted by the Random Forest model, and the stock market volatility is predicted by the usual REGARCH and RSVL models. Based on the research of this article, I hope it will be helpful to relevant researchers in the study of stock market forecasts in the future.