Investor Sentiment Mining Based on Deep Machine Learning and Its Impact on Stock Price Crash Risk
In order to explore whether the emotional tendencies contained in the text of the Internet stock bar community post will affect the stock price crash risk,this paper builds a convolutional neural network and long and short memory neural network feature fusion model(LSTM-CNN)to identify and classify the real-time posting text of sample stocks in Oriental Fortune Stock Bar,and constructs monthly indicators of investor sentiment.Empirical research has found that investor sentiment in the current period has a significant positive impact on the risk of future stock price crash,and the more optimistic the investor sentiment is,the greater the risk of a stock price crash.Under different market environments,emotions have an asymmetric impact on the stock price crash risk,with a more significant positive impact of emotions on the risk of crash in bear markets.Further heterogeneity analysis indicates that investor sentiment has a more significant impact on the stock price crash risk in sample companies with smaller scales,lower equity concentration,greater short selling restrictions,and lower marketization levels in the company's location.In addition,we find that stock liquidity is an important mediating variable for investor sentiment to affect the stock price crash risk.These conclusions may explain the formation mechanism of stock price crash risk from the perspective of investor sentiment,enriching the understanding of the factors influencing stock price crash risk.
万方数据
维普
