Applicability of Financial System Using Deep Learning Techniques

Abstract

To predict the future stock price is not a slipshod task due to the unpredictable behavior of stock market. The financial market always tends to be a great challenge for experts to predict the future stock prices. Hence, market experts use varied differential techniques such as fundamental or technical to determine the financial forecasting. In past, there exist various regression models, such as ARIMA, ARCH, and GARCH which are used for technical analysis, but they tend to be less robust for non-stationary data. Moreover, varied machine learning algorithms are applied in the past which include SVM and random forest for financial forecasting. However, results retrieved from these algorithms are still questionable with time series data. In the current study of approach, we have utilized deep learning models which have a tendency to be more focused on right features by themselves; it requires very little intervention of developer. Hence, deep learning methods can increase the ability of technical analysis for future stock prediction. In this paper, varied comparison is drawn among the two deep learning approaches such as LSTM and CNN which are applied to predict the future stock prices of Infosys company. Empirically, it has been proved that the LSTM outperforms the CNN.