Stock market prediction using neural network through news on online social networks

Abstract

Stock market prediction has attracted a lot of attention from both business and academia. In this paper, we implement a model based on Recurrent Neural Networks (RNN) with Gated Recurrent Units (GRU) to predict the stock volatility in the Chinese stock market. We also propose many price related features which are used as inputs for our model. Apart from that, we carefully select official accounts from Chinese largest online social networks - Sina Weibo and extract the content posted by these accounts to analyze the public moods. An influence feature is derived based on the public moods to further improve the prediction model. The experimental results show that our model outperforms the baseline method and can achieve a good prediction performance.