Comparative Analysis of Regressor Machine Learning Models in Forecasting SPDR S&P 500 ETF Trust (SPY) Movements

Stock market forecasting remains one of the most challenging tasks in finance due to the market's high volatility, complex dynamics, and sensitivity to external events. This study investigates the short-term predictive performance of five regressor models on the S&P 500 ETF Index (SPY). There are two experimental setups for evaluating models: one incorporating technical indicators and one excluding them. The aim is to determine whether technical indicators enhance prediction accuracy and to identify which model is most effective for short-term forecasting. Also, the model's performance is assessed using MAE, RMSE, R2 score, and directional accuracy. The results showcase that the R2 score is a poor indicator in short-term financial datasets. Directional accuracy with technical indicators ranged from 51-54%, a result better than randomly guessing. Without technical indicators, models ranged from 45-48%, highlighting the importance of technical indicators in predictions.