This study examines the effectiveness of incorporating investor sentiment into machine learning models based on decision trees-specifically, Random Forest, XGBoost (Extreme Gradient Boosting), and LightGBM (Light Gradient Boosting Machine)-for option pricing in the Taiwan market. The empirical results demonstrate that these machine learning models significantly outperform the traditional Black-Scholes model in pricing accuracy. Notably, adding investor sentiment enhances the models' pricing performance, especially for at-the-money and in-the-money options, where pricing errors are reduced by 4 times and 2.6 times, respectively. The Random Forest model exhibits the best performance overall.
Cheng et al. (Fri,) studied this question.