The fluctuation of stock prices affects various aspects of social life, and thus, effective prediction of stock market trends has significant commercial and social value. Generally, predicting future market trends depends on studying and extracting the changing patterns of historical data. In this study, based on 1000 trading days from January 2018 to December 2022 in the Chinese stock market and 6000 stock comments, we propose an LSTM model based on historical financial data and investor sentiment to predict the future trends of the Chinese A-share market index. Firstly, we extend the sentiment lexicon to fit the context of stock market and construct specific matching rules to calculate sentiment scores. Secondly, we construct an LSTM model based on financial data, and compared the prediction results with the traditional ARMA model with the same inputs, where the empirical evidence cannot support the argument that the LSTM models are superior in predicting performance. Finally, we add the sentiment index to the input layer of the LSTM model, and compare the prediction results of the new model and the basic model on the test set, fully showing the significant role of sentiment indicators in predicting market trends.