Machine learning-driven factor fitting model for stock data and its future trend prediction

Stock data factor model fitting and forecasting has been a hot topic in fintech and quantitative investment research. As a representative technology of artificial intelligence, machine learning can greatly improve the effect of forecasting research in economics and management. This paper aims to analyze stock data and predict fu-ture trends by using a variety of factor fitting models. First, we ex-tract multiple related factors from the data, such as volatility factor, growth factor, momentum factor, size factor, value factor, liquidity factor, profit factor, etc. Then, we use these factors to build fitting models, in this paper, eight machine learning algorithms, including linear model, Lasso regression, Ridge regression, decision tree model, random forest model, GBDT model and XGBT model, are used to build stock return prediction model and investment portfolio. Through these models, we can make pre-dictions about the stock data and come up with future trends. Final-ly, through empirical analysis, we verify that the forecast of these factor fitting models is better than that of CSI 300, and the annual-ized return rate and Sharpe ratio are both higher than that of CSI 300. Among them, the GDBT model has the best forecast results, with Sharpe ratio reaching 1.07 years and annualized return rate reaching 0.92, far higher than the 0.15 annualized return rate of CSI 300.