Sentiment-based stock price prediction in developing countries: Evidence from Iran

This study examines the predictive ability of public (investor) sentiment from web financial news and social media within the Tehran Stock Exchange (TSE) between November 1, 2020, and July 31, 2022. The study examines the influence of Persian-language sentiment data on movements in stock prices across various time horizons by employing a fine-tuned ParsBERT large language model architecture and a Support Vector Machine to predict the stock price movement over 1-day, 7-day, 30-day, and 90-day intervals. Findings show that the forecasting performance of public sentiment from web financial news exceeds that of social media in short-term predictions; however, social media sentiment provides more reliable and accurate projections for medium- to long-term intervals. Furthermore, combining sentiment data and technical indicators significantly enhances forecast accuracy. These findings highlight the unique role of sentiment analysis in understanding investor behavior in politically influenced and developing economies.