STOCK MARKET PREDICTION VIA SENTIMENT ANALYSIS OF FINANCIAL AND ECONOMIC NEWS USING MACHINE LEARNING

Abstract

This study investigates the use of sentiment analysis derived from financial news articles to predict stock market returns. By incorporating sentiment features along with traditional stock market predictors, we aim to enhance prediction accuracy for stock price movements. Various machine learning models, including Linear Regression, Decision Trees, Random Forests, Gradient Boosting Machine (GBM), and Extreme Gradient Boosting (XGBoost), were evaluated for their performance in predicting stock returns. The models were assessed using several regression metrics, such as RMSE, MAE, MAPE, and R², with a particular focus on how well sentiment-based features, like SentNews and SentNewsIBOV, contribute to predicting market movements. The study also highlights the importance of identifying the most influential words within financial news articles using feature importance analysis, particularly through the XGBoost model. The results indicate that sentiment data, especially when combined with advanced machine learning algorithms like XGBoost, provides a valuable tool for improving the accuracy of stock return forecasts. The findings suggest that news sentiment, which reflects public perception and market sentiment, significantly influences stock price movements and can be leveraged to forecast future market trends.