Improving Accuracy in Market Close Prediction Using Optimized Financial Indicators

The stock market is a dynamic and complex financial system where buyers and sellers trade shares of publicly listed companies. Predicting next-day stock prices involves understanding the interplay of multiple factors that influence price movements. This study employs various machine learning algorithms to predict next-day stock prices using a comprehensive set of features, including 27 indicator values, along with metrics such as open, close, high, low, volume, and the overall market index value. A total of 17 machine learning algorithms and nine feature selection techniques were utilized, with models evaluated using RMSE, MAE, and MAPE. Unlike previous studies, which often focus on price prediction using limited combinations of features and algorithms, this research adopts a novel approach by integrating these diverse elements. Additionally, the motor sector, an underexplored area in stock price prediction, was analyzed. Among the models tested, Ridge regression with forward feature selection demonstrated the best performance, achieving a MAE of 0.881. This study lays the groundwork for future research by extending the scope to other sectors and exploring advanced machine learning and deep learning models combined with feature selection techniques.