Deciphering news sentiment and stock price relationships in Indonesian companies: an AI-based exploration of industry affiliation and news co-occurrence

Abstract

Abstract The rapid increase of textual data has transformed the way we understand and forecast financial market behavior. Investor sentiments, often swayed by news, are pivotal in determining stock prices. Analyzing a dataset of 192.582 Indonesian financial news articles published between 2018 and 2023. This study investigates the complex connections between news sentiment and stock market behavior of Indonesian companies. We leverage AI-based sentiment analysis and natural language processing techniques, including identity recognition, network analysis, and correlation assessment, to explore how news sentiment affects stock prices at the levels of individuals, industries, and news co-occurrence clusters. While earlier research has addressed the effect of sentiment on stock prices at both the company and industry levels, there is a significant lack of studies focused on media co-occurrence clusters, which is vital for comprehending the collective media portrayal of interconnected firms. Our results show that sentiment-price correlations strengthen hierarchically, with individual companies at 0.26, industry groupings at 0.30, and news co-occurrence clusters at 0.43. This research introduces a unique analytical framework that explores sentiment across various levels, highlighting co-occurrence clusters that reflect business relationships beyond traditional industry lines. It demonstrates that companies frequently mentioned together in the news exhibit stronger and more stable sentiment-price correlations, offering a new analytical perspective for AI-driven investment strategies and underscoring the potential of big data analytics in Indonesia's capital market.

Keywords

• AI-driven sentiment analysis
• Stock market dynamics
• Entity recognition
• News co-occurrence
• Indonesian companies