Enhancing Stock Price Prediction: An Ensembled Time Series Forecasting

The financial market is a dynamic environment, which complicates accurate forecasting. Forecasting financial markets is essential for informed investment decisions and effective risk management. The proposed work seeks to establish a robust forecasting framework that employs technical indicators in ensemble learning techniques to increase prediction accuracy. Indicators are generated to identify market patterns and trends. The research uses an ensemble learning methodology that integrates aggregated decision trees, additive boosting, and support vector regression (SVR). The ensemble forecasting model was implemented on banking stocks listed on the National Stock Exchange (NSE) over various time frames: 1 day,5 days, and 10 days, and 21 days. Compared with existing models, the proposed model demonstrated significant improvement in prediction, as evidenced by the reduction in the mean absolute percentage error (MAPE) across all time horizons. For 1 day, the MAPE was reduced from 0.0009 to 0.2336. Similarly, 5 days showed improvements from 0.0021 to 0.2339. The 10-day MAPE decreased from 0.0065 to 0.2368, and the 21-day MAPE improved from 0.0065 to 0.2277. The findings underscore the efficacy of the Ensembled Time Series Forecaster in leveraging the strengths of multiple machine learning models. This synergy enhances forecasting precision, making it a robust tool for financial analysts and investors in navigating complex stock market behaviors. The adaptability of the proposed method across short- and long-term horizons makes it a valuable asset for real-world financial decision-making.