Efficient Stock Price Prediction Strategies: A Multi-Model Analysis in the Indonesian Capital Market

Abstract

Accurate stock price prediction plays a vital role in finance, influencing investment strategies, market efficiency, and overall economic stability. The Indonesian capital market has experienced rapid growth, with increasing trading volumes and greater retail investor participation. This study investigates the efficacy of a hybrid model combining Random Forest (RF), Gated Recurrent Unit (GRU), and Seasonal Autoregressive Integrated Moving Average with Exogenous factors (SARIMAX) for forecasting stock prices in this dynamic market context. Using stock price data from Indonesia’s top five banking institutions, sourced via the Yahoo Finance API, each model (RF, GRU, and SARIMAX) was first developed independently and then combined through an ensemble approach. The goal was to evaluate whether this integrated method could yield more accurate predictions than standalone models. Results indicate that the hybrid model, particularly the RF-GRU-SARIMAX combination, outperforms individual models, delivering notable improvements in predictive accuracy. Random Forest demonstrated reliable and consistent performance across all bank stocks, while GRU exhibited limitations in accuracy, and SARIMAX, as a conventional statistical model, showed relatively weak performance in price forecasting. This study highlights the importance of using adaptive predictive models like Random Forest to enhance stock price forecasts in emerging markets. Recommendations for further research include refining the GRU model, exploring additional hybrid combinations, applying robust validation techniques, and incorporating more exogenous variables. Empirical evidence from this research offers valuable insights for investors and policymakers, supporting improved decision-making within the Indonesian capital market.