From the perspective of behavioural finance, this paper combines the fine-grained sentiment calculation with the stock market econometric model to explore the interactions between netizens’ sentiments and stock returns, analyze the differences in the influences of various emotions expressed by netizens on the stock market. First, it constructs a sentiment dictionary for the financial field; then, it calculates the emotion values contained in the text corpus, and constructs a textual sentiment classifier based on the recurrent neural network, calculates the emotion value and establishes the daily netizen sentiment index; and finally, it builds an econometric model to study the interactions between the netizen sentiment index and the stock returns. The results show that this model improves the accuracy of sentiment classification, reduces the number of iterations and saves computing resources; and that the netizen sentiment index, especially, “disgust” and “like”, has significant effects on the stock price changes and transaction volumes, while on the other hand, the listed company’s stock returns data has no reverse effect on the netizen sentiment index.
Zhu, Meng-xuan; Zhang, Wei; Wang, Meng; Kong, Sui-xi; and Wang, Jian, "Using Fine-grained Emotion Computing Model to Analyze the Interactions between Netizens’ Sentiments and Stock Returns" (2019). WHICEB 2019 Proceedings. 20.