The rapid development of Internet has enabled non-professional individual investors to share information and express their tendency through online financial community. Using 5 178 824 user comments on eastmoney. com，this paper uses the convolutional neural network (CNN) classification algorithm to extract and measure the support tendency of online users for the future market (i. e. bullish or bearish) and the impact of user support tendency on the market is tested from two aspects of market returns and volatility. The results show that the current user’s support tendency has a significant negative impact on future stock market returns，and the consistency of support tendency will amplify the market’s volatility. Furthermore，the user support tendency is significantly based on the historical performance of the stock market，and the support tendency has a certain“herd effect”.