Binarized neural network with depth imaging techniques for stock return direction prediction

This research article proposes a new robust mechanism for stock return prediction using Binarized Neural Network model. Some of the key bottlenecks in deep learning models include higher memory footprint requirements and extensive computational capacity. We have addressed these bottlenecks and have considerably reduced the same using binarized direction prediction algorithm for forecasting the direction of the stock market index. In this work, ten years of stock data listed under the Indian stock market NIFTY-50 has been considered with three different imaging techniques to create dataset and analysis. The proposed model has been designed to forecast the direction of the stock market index with a reduced memory footprint without using any high performance computing systems. The binarized neural network with durational feature based depth imaging technique for direction prediction achieved 95.916% accuracy which is equivalent to its full precision counterpart 95.98%. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.