Deep learning CNN model-based stock price prediction by candle stick images

An efficient technique for the trading prediction of stock market price is critically challenging, especially in the real time prediction for the traders. Development of a real time effective algorithm is still under development by the various trading support companies to increase profit from the correct prediction. In this paper focus on the development prediction method using deep learning techniques for the market by candle stick chart generation from the previously available dataset. Convolution neural networks are used for training candle chart image and prediction for low or high closing on next day for the given candle chart of any given window size. In the proposed CNN-based models result, particularly VGG19, achieved a high testing accuracy for the S&P500 dataset the VIX Index using 20-candle window images, demonstrating the better performance compared to ResNet50.