A Multi-factor Approach for Stock Price Prediction by using Recurrent Neural Networks

Xu Zhang, Chen Li, Yasuhiko Morimoto


Stock price prediction is a difficult type of time series predictive modeling problem. In time series forecasting, Autoregressive Integrated Moving Average(ARIMA) is one of the famous linear models. However real-world time series like the stock is rarely pure linearity. The conventional ARIMA method cannot model the non-linear function very well. Since the stock price forecast depends on many factors, generating a good prediction model is a huge challenge for the researchers. In recent years, Recurrent Neural network (RNN) has yielded immense success on time-series prediction which can learn arbitrary linear and non-linear function from the dataset. Long Short-Term Memory network (LSTM) is one of RNN which is usually used for solving gradient vanishing problem. In this paper, we propose an LSTM approach to predict stock price. Moreover, for better improving the accuracy of our model, we consider multi-factor which are relevant to stock price. In the experiments, we firstly compare our model with the conventional ARIMA method. Besides, we also demonstrate the performance of single-factor with the multi-factor model and different time periods.


Stock Price; Recurrent Neural Network; ARIMA

Full Text:



  • There are currently no refbacks.