Ensemble Regression Techniques for Enhanced Stock Price Forecasting

Stock Price Prediction has been a concern for many financial analytics and business owners. This paper focuses on a machine learning framework for predicting stock prices accurately and comprehensively. The dataset used in this paper contains 365 daily records of five stocks. The data provides 218 characteristics, such as technical indicators, lag factors, and statistical measurements. Feature reduction technique like Principal Component Analysis (PCA) is used in this paper for predicting accuracy by utilizing both variance-based and fixed-component approaches. A variety of regression models and boosting techniques have been used in this paper such as Random Forest, XGBoost, Linear Regression, and Support Vector Machines. This paper uses a time-aware data partitioning and cross-validation techniques to guarantee accurate performance predictions. The results clearly show that SVR Linear outperforms other algorithms in terms of predicting stock prices accurately with R2 scores of 99.97% and RMSE values as low as 0.0585.