Sentiment-Driven Forecasting of the Stock Index in Vietnam: A Machine Learning Perspective

Abstract

This study investigates whether news sentiment, when combined with technical indicators, can enhance the prediction of directional movements in the Vietnamese VN-Index. By constructing a novel dataset of 6,480 Vietnamnet news articles and integrating it with historical market data from 2019 to 2025, the research applies a suite of machine learning classifiers—including Logistic Regression, SVM, Random Forest, Gradient Boosting, CatBoost, AdaBoost, and Naive Bayes—to classify index changes as either "Up" or "Unchanged/Down." The models are evaluated using 10-fold cross-validation and assessed across accuracy, precision, recall, F1-score, and ROC-AUC metrics. Results show that ensemble-based models, particularly CatBoost and Gradient Boosting, outperform linear and probabilistic baselines, confirming that sentiment-derived features add meaningful predictive power when combined with traditional technical indicators. The findings contribute to the literature on behavioral finance and machine learning by demonstrating that investor sentiment, as reflected in news narratives, plays a significant role in shaping short-term stock index movements in emerging markets.