Analysis and Forecasting of Financial Time Series Using CNN and LSTM-Based Deep Learning Models

Abstract

AbstractDesigning predictive models for forecasting future stock price has always been a very popular area of research. On the one hand, the proponents of the famous efficient market hypothesis believe that it is impossible to accurately predict stock prices, on the other hand, propositions exist in the literature that demonstrate that it is possible to very precisely predict stock prices by accurate modeling of the predictive systems. We propose a robust framework consisting of a suite of deep learning-based regression models that yields a very high level of accuracy in forecasting of stock prices. The models are built using the historical stock price data of a well-known company listed in the National Stock Exchange (NSE) of India. The stock prices are recorded at five minutes interval during the period December 31, 2012 to January 9, 2015. Exploiting the features in these extremely granular stock price data, we build four convolutional neural network (CNN) and five long- and short-term memory (LSTM)-based deep learning models, validate them and finally test them on their performance. Extensive results are presented on the performance of the models on two metrics- execution time and root mean square error (RMSE) values.KeywordsStock price predictionRegressionLong and short-term memory networkWalk-forward validationMultivariate time series