Stock Trend Prediction by Fusing Prices and Indices with LSTM Neural Networks

Abstract

Forecasting stock market prices and trends is a promising area of machine learning. In the present paper we focus on the application of deep learning, a promising category of technical analysis, and in particular LSTM neural networks. In the context of this paper, stock market forecasts are estimated for large technology companies and the factors affecting their performance are studied. Stock market predictions can significantly benefit from the fusion of different information sources providing insights of diverse types and levels. Here we propose the use of LSTM neural networks in order to fuse different types and levels of information and generate predictions about the future prices and the future stock price trend. Specifically, the proposed LSTM model takes as input the price of the stock under examination along with related indices. Hence, it considers not only the historical price data but also the indices implying changes in the market due to several external influences. Our results demonstrate that the proposed approach outperforms other approaches which consider only the historical price data in terms of the price movement prediction.