Stock trend prediction based on fusion of alternative text and multiple features

Stock price prediction has always been a prominent research field. How to predict stock price more accurately is very important for financial research. In this paper, the Shanghai Composite Index of 2020 is taken as the data set and the alternative text represented by news public opinion is used to predict the overall stock trend according to the efficient market hypothesis. This model crawls the news text of a specific financial website, mines the emotional information contained in the text with Word2Vec and LSTM, and scores the public opinion on a specific date. Then, the traditional stock price features are combined with the public opinion scoring features to form a multi-feature input matrix. Finally, the CNN-LSTM neural network is used to predict the future stock trend with high accuracy. The experimental results reveal that the suggested model has a considerable improvement in accuracy when compared to standard time series models and machine learning models.