Stock Market Prediction Based on Historic Prices and News Titles

Predicting the trend of stock market prices is a very challenging task, in that stock markets are complicated and can be influenced by a variety of factors. Despite the great difficulty, predicting the trend of stock market prices accurately is very meaningful and can bring a large amount of profit. In the past several decades, a lot of studies have been done on this problem. But most of the methods take only the historic prices data as the input, which is not enough for such a complicated problem. In this paper, a hybrid method taking both historic prices and the news as input is proposed. The hybrid model combines the best of two kinds of networks--RNN-LSTM for time series data and CNN for abstract high-dimensional data. These two different kinds of networks are combined together to make a prediction. A set of experiments have been carried out to show the performance of the proposed method. The result obtained is promising, and the propose method achieves a great degree of accuracy in prediction and outperforms the baselines a lot.