Stock Return Prediction Based on Random Forest

Abstract

With the deepening application of machine learning technology in the financial sector, using algorithms to capture non-linear market patterns has become a significant direction in quantitative investment. This study aims to utilize the Random Forest algorithm from ensemble learning to predict the short-term returns of representative stocks across multiple markets. The research acquired daily trading data for the past five years for targets such as Kweichow Moutai, Contemporary Amperex Technology (CATL), and Apple Inc. (AAPL) from Yahoo Finance (yfinance) and AkShare. Feature variables comprising technical indicators and volatility factors were constructed, with the subsequent five-trading-day return as the prediction target. Using Python's Pandas library for data cleaning and feature engineering, the Random Forest model was built based on the Scikit-learn library and evaluated using metrics such as MSE, MAE, and Directional Accuracy. Backtesting and visual analysis results indicate that the Random Forest model possesses a certain level of feasibility and application potential in short-term return prediction, offering investors a data-driven and objective decision-support tool. This study also discusses the model's limitations and proposes directions for future improvement.