Innovative Deep Learning Model-based Stock Price Prediction using a Hybrid Approach of CNN and Gradient Recurrent Unit

Abstract

The fluctuation in stock prices from industry to industry are a major source of concern in the market. As the market attracts more participants and stock prices play a larger role in more transactions, the ability to accurately predict stock price movements becomes more valuable. When making an investment, many people first look at the share price and then try to anticipate whether or not that price will go up or down in the future. The traditional problem of forecasting the stock market using Machine Learning tools and methodologies has been thoroughly studied. Time dependence, volatility, and similar complicated dependencies are interesting aspects that make this modeling non-trivial. To overcome this the proposed method in this work is a deep learning-based hybrid strategy for predicting stock prices. Preprocessing is performed after input is delivered to increase precision. Selecting MPSOA features is the next step. In the end, it's put to use in MC-GRU model training. The proposed method achieves better results than both the CNN and GRU models.