MACHINE LEARNING APPROACHES TO DIVIDEND PREDICTION: AN EMPIRICAL STUDY ON VIETNAM’S STOCK MARKET

In finance, classifying dividend payouts has garnered considerable attention due to its impact on investment decisions. While prior studies have primarily explored the legal and policy dimensions of dividend distribution, this paper centers on analyzing financial indicators to categorize dividend payouts. Most existing research has concentrated on the legal and regulatory frameworks that influence these payouts. In contrast, our study focuses exclusively on identifying the financial factors that affect dividend payout decisions. To enhance the robustness and reliability of our analysis, we implemented various data processing techniques, including managing missing values, detecting outliers, addressing data imbalances, and applying k-fold cross-validation. Our comprehensive approach involved testing 10 different models, such as KNeighbors Classifier, Decision Tree Classifier, Random Forest Classifier, Gradient Boosting Classifier, XGB Classifier, LGBM Classifier, CatBoost Classifier, ExtraTrees Classifier, AdaBoost Classifier, and Bagging Classifier. Among these, the ExtraTrees Classifier emerged as the top performer, achieving the highest accuracy at 81.61%, with a recall of 80.18%, an F1 score of 80.69%, precision of 82.52%, and Cohen’s Kappa of 63.58%. The Random Forest Classifier also showed strong performance, with an accuracy of 79.33% and Cohen’s Kappa of 58.58%. These findings demonstrate the efficacy of machine learning algorithms in accurately classifying dividend payouts. The resulting binary classification system is highly valuable for investors making informed decisions regarding investment opportunities, risk management, and portfolio diversification.