This study aims to investigate the predictive power of various financial indicators in forecasting daily stock returns.To achieve this goal,we collected a year’s worth of stock data for the stocks listed in S&P 5...This study aims to investigate the predictive power of various financial indicators in forecasting daily stock returns.To achieve this goal,we collected a year’s worth of stock data for the stocks listed in S&P 500 using yfinance,which included market capitalization,sector,PE ratios,revenue,earnings per share,and other variables.Our methodology involved analyzing stock performance predictors using advanced statistical methods and machine learning models,including Random Forest,Gradient Boosting,and Generalized Linear Model Regression.The findings suggest that daily stock returns can be significantly predicted by a few financial factors including PE ratio,PB ratio,and Debt to Equity ratio.To assess the performance of our models,we performed a 12-fold cross validation on the validatable models Gradient Boosting and Random Forest.The result demonstrated that Gradient Boosting provided the most reliable predictions based on mean squared error analysis.This study validates the robustness of using historical stock information in forecasting short-term stock returns,providing insights that could benefit investors in understanding market dynamics.展开更多
文摘This study aims to investigate the predictive power of various financial indicators in forecasting daily stock returns.To achieve this goal,we collected a year’s worth of stock data for the stocks listed in S&P 500 using yfinance,which included market capitalization,sector,PE ratios,revenue,earnings per share,and other variables.Our methodology involved analyzing stock performance predictors using advanced statistical methods and machine learning models,including Random Forest,Gradient Boosting,and Generalized Linear Model Regression.The findings suggest that daily stock returns can be significantly predicted by a few financial factors including PE ratio,PB ratio,and Debt to Equity ratio.To assess the performance of our models,we performed a 12-fold cross validation on the validatable models Gradient Boosting and Random Forest.The result demonstrated that Gradient Boosting provided the most reliable predictions based on mean squared error analysis.This study validates the robustness of using historical stock information in forecasting short-term stock returns,providing insights that could benefit investors in understanding market dynamics.
