SENTIMENT ANALYSIS AND ENTITY DETECTION ON NEWS HEADLINES TO SUPPORT INVESTMENT DECISIONS

Accurate investment decisions are often influenced by information available in the media. News headlines, as part of information media, can provide an initial picture of market sentiment and ongoing trends. This research examines the importance of making appropriate investment decisions with a focus on sentiment analysis and entity detection in news headlines as supporting tools. Through machine learning-based sentiment analysis and Named Entity Recognition (NER) techniques, this study identifies opinions and entities such as company names, stock indices, and industry sectors in news headlines. This research compares three machine learning algorithms, namely SVM, Naive Bayes, and Random Forest using cross-validation. The result shows that the best algorithm is SVM with weighted average F1-score of 76,68%. Furthermore, hyperparameter optimization is performed using Optuna for the SVM algorithm, which is an innovation in the context of sentiment analysis on news headlines in Indonesia. The result shows an increase in weighted average F1-score to 78,14%. For NER, a rule-based method is used by utilizing the Jaro-Winkler string similarity function. The combination of sentiment analysis and NER is then presented in the form of a dashboard using Google Looker Studio tools, with data from sentiment analysis and NER results being processed periodically and automatically using Google Workflows. This research makes a significant contribution by expanding the scope of analysis from just one or a few issuers to all entities published on news portals thanks to NER support, making the results relevant to support investment decisions that are responsive to dynamic market changes.