The Effect of News Sentiment on Jakarta Composite Index Prediction Using Support Vector Regression Method

This research investigates the impact of news sentiment on predicting the Jakarta Composite Index (JCI) using the Support Vector Regression (SVR) method. Market sentiment, derived from news articles, has been analyzed to understand its influence on stock price movements. A dual dataset approach was employed, consisting of financial news articles from Kompas.com and historical JCI stock data. The research incorporates sentiment analysis using ChatGPT large language models (LLMs), which are then integrated as features into the prediction model. Five scenarios of sentiment integration were evaluated to identify the most effective approach. The results indicate that Scenario 4 consistently delivers the highest prediction accuracy across different evaluation metrics, with Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values of 0.009555 and 0.007298 in the log metric evaluation, 47.616867 and 36.52605 in the absolute metric evaluation, and 85.146387 and 70.34775 in the stock closing price evaluation. While sentiment integration shows potential, its success is scenario-dependent and influenced by hyperparameter tuning. This research underscores the utility of sentiment analysis in enhancing stock price predictions and provides a foundation for further exploration of sentiment based predictive models in financial markets.