Media and Investor Sentiment Prediction for Stock Returns - A BP Neural Network Model based on Whale Optimization Algorithm

Currently, behavioral finance theory has incorporated investor psychology into the research framework of securities. Most investors in China's capital market are retail investors, and it is relatively difficult for them to acquire comprehensive stock information and professional financial knowledge. Therefore, they are more likely to be influenced by external public opinion and their own emotions. Investor sentiment plays a significant role in predicting stock returns. This paper is based on text sentiment analysis of investor comments obtained through web crawling and reports about the stock market from various financial newspapers and local media. By using a sentiment dictionary for sentiment analysis, sentiment scores are assigned to the texts, eventually forming two variables: industry sentiment index and individual stock sentiment index, which are used as the basis for predicting the daily return rate of stocks. We select and construct a whale optimization algorithm to optimize the BP neural network. By analyzing the accuracy and practicality of the BP neural network model in predicting stock returns considering both industry sentiment and individual stock sentiment, we found that the BP neural network optimized by the whale algorithm has a high accuracy in stock market prediction. The research result provides approach ideas for predicting the returns of a single stock in a specific industry. It also offers new directions for the application of media texts and stock reviews. Additionally, it expands the scope of reasonable suggestions for investors in stock selection and stock operations.