Multi-sentiment fusion for stock price crash risk prediction using an interpretable ensemble learning method

With the development of the Chinese security market, stock price crashes have occurred frequently, and extant studies have proved that investor sentiment is one of the most significant factors that have led to dramatic volatility of stock prices. Additionally, irrational sentiment poses formidable challenges to traditional methods of predicting stock price crashes. Based on an ensemble learning approach, this research endeavors to explore the effects of investor sentiment indicators on stock price crashes within the Chinese stock market. First of all, several advanced machine learning techniques are employed for stock price crash predictions. Meanwhile, the hyperparameter optimization method is applied to improve the performance of the prediction model. In addition, the interpretable machine learning method is applied to analyze the important investor sentiment indicators that significantly influence stock price crashes. Experiment results show that the method proposed in this research performs better than the benchmark methods, especially for the small-cap samples. Additionally, foreign investor sentiment indicators contribute more to prediction results than individual and institutional sentiments. The proposed method can provide market regulators and investors with an innovative perspective on stock price crash prediction with multiple sentiment fusion.