The impact of correlation-based feature engineering on ensemble model

Abstract

Random Forest is a representative ensemble learning algorithm whose core concept lies in integrating multiple decision trees to enhance prediction accuracy and model stability. This study takes the Chinese stock market as the research background, selecting six representative stocks from the pharmaceutical and automotive industries as research subjects. The Random Forest algorithm is applied to predict stock closing prices. By employing feature selection and data preprocessing, the model’s generalization capability and predictive accuracy are improved. Experimental results demonstrate that Random Forest effectively captures nonlinear characteristics in financial data, providing a feasible technical approach and valuable reference for forecasting stock price trends across different industries.