Prediction of Stock Market using Stochastic Neural Networks

The primary objective of investors and stockbrokers is to make profits by being able to predict the financial markets. However, forecasting is a complex task since the financial markets have a complicated pattern. This study addresses the direction of the stock price index for Japanese Nikkei 225. The research compares two prediction models, i.e., the Stochastic Neural Networks (SNN) and fusion of Long -Short Term Memory and Stochastic Neural Networks (LSTM - SNN) for predicting the index. The input layer includes computation of fifteen technical indicators using stock market parameters (open, high, low, close prices, and volume). Accuracy of each of the prediction models was evaluated using price and trend performance metrics. The evaluation was carried out for historical data from 23rd January 2007 to 30th December 2013 of the Tokyo Stock Exchange (TSE). The experimental outcomes recommend that for the SNN, the model gave an accuracy of 85.37% and hybrid of LSTM – SNN gave accuracy of 86.28%. The increase in the accuracy of LSTM – SNN was due to the introduction of LSTM layer. Experimental outcomes also illustrate that the performance of both the prediction models progress when these technical indicators are added to the input layer of the proposed models.