Ensemble Model to Forecast Sri Lankan Stock Prices Using Macroeconomic Variables

Abstract

"The stock market plays a critical role in Sri Lanka’s economy, serving as a platform for businesses to raise capital and for investors to grow wealth. However, due to the inherent risks of stock market investments, making well-informed decisions is essential to maximize returns. Economic instability and its influence on macroeconomic factors further complicate the decision-making process, leaving many investors hesitant to engage in stock market activities. This hesitation is largely driven by the difficulty of determining which stocks to invest in, highlighting the need for effective tools that support informed investment choices.

This research seeks to address this challenge by developing an ensemble model capable of predicting stock prices for Sri Lankan companies based on macroeconomic indicators. By combining forecasts from multiple individual models, the ensemble approach aims to improve prediction accuracy and reliability.

The model is developed and validated using data from the Colombo Stock Exchange and the Central Bank of Sri Lanka. The study evaluates seven machine learning models, ultimately integrating four high-performing ones—single-layer LSTM, Bi-directional LSTM, GRU, and RNN—into the ensemble. These models were chosen based on their ability to achieve a lower Root Mean Square Error (RMSE). The proposed ensemble model achieves an RMSE of 4.71, reflecting its predictive accuracy.

The successful application of this ensemble model can provide investors with critical insights into stock market trends, enabling more confident and informed investment decisions. This research not only addresses a pressing need for better financial forecasting tools in Sri Lanka but also contributes to the broader fields of financial analytics and machine learning by exploring the use of ensemble methods for stock market prediction."