Exploring Cluster Stocks based on deep learning for Stock Prediction

Abstract

Stock prediction is an important yet challenging problem in finance field. As the stock increase, cluster stocks have a certain on target stock, because the same type of stocks have a linkage effect. However, a large amount of approaches only utilize their own information, and not use extra information from cluster stocks. In this paper, to improve the stock prediction, we exploit cluster stocks, and develop a deep learning model for stock prediction. A LSTM model is first introduced, then we introduce attention model. The combination LSTM and attention model is proposed for stock prediction by using cluster stocks. To validate the proposed approach, we collect the data from Wind Financial Terminal. Extensive experiments are carried out on this data, and results show that exploring cluster stocks based on deep learning is a promising development that improves the performance of stock prediction, and the cluster stocks information can offer promising enhancement.