Predicting Tehran Stock Exchange Performance: A Focus on Trading Volume and Ensemble Learning Models

In this study, we investigate the effectiveness of ensemble learning models in predicting stock market movements on the Tehran Stock Exchange (TSE). The unique regulatory environment and dominance of government-affiliated shareholders in the TSE present distinct challenges and opportunities for predictive modeling. We employ Random Forest, LightGBM, XGBoost, and CatBoost algorithms, considering both technical indicators and trading volume features as predictors. Our analysis reveals that while all models demonstrate strong predictive performance, Random Forest and LightGBM exhibit slightly superior accuracy, precision, and recall compared to XGBoost and CatBoost. Furthermore, feature importance analysis highlights the significance of trading volume indicators in predicting stock market trends on the TSE, suggesting their potential utility in enhancing predictive models for similar market conditions. Our findings underscore the importance of model selection and feature engineering in developing robust predictive frameworks for stock market forecasting, with implications for investors, financial analysts, and policymakers.