MACHINE LEARNING MODELS FOR STOCK PRICE PREDICTION: A COMPREHENSIVE REVIEW OF DSE APPLICATIONS

Abstract

The prediction of stock prices remains a complex challenge due to the non-linear and volatile nature of financial markets. Traditional methods, such as fundamental and technical analyses, struggle to capture these complexities, limiting their effectiveness. This study examines the application of machine learning (ML) techniques to improve stock price forecasting, focusing on the Dhaka Stock Exchange (DSE). The research aims to explore how ML models, including supervised learning, deep learning, and hybrid approaches, can enhance prediction accuracy by identifying complex patterns in large datasets. The study reviews various methodologies, including Long Short-Term Memory (LSTM) networks, Support Vector Machines (SVM), and ensemble models, alongside hybrid models that combine artificial neural networks (ANN) with technical indicators like Moving Average Convergence Divergence (MACD) and Relative Strength Index (RSI). Time series data from the DSE is used in these studies to evaluate prediction performance. The results show that ML techniques significantly outperform traditional methods in forecasting short-term stock price trends. Hybrid models, particularly those integrating ANN with technical indicators, offer higher precision. The incorporation of sentiment analysis and big data analytics further improves model adaptability to dynamic market conditions. The major findings indicate that while ML models enhance prediction accuracy, challenges such as the limited availability of high-quality datasets, the lack of integration of macroeconomic factors, and difficulties in real-time validation remain. The DSE’s high volatility and sectoral variability further complicate accurate predictions.

JEL Classification Codes: H54, P42, G17, C88.