国际群体智能与进化计算杂志

抽象的

Artificial Neural Networks Based Indian Stock Market Price Prediction: Before and After Demonetization

Siddheshwar Chopra*, Dipti Yadav and Anu Nagpal Chopra

In this paper, stock market price prediction ability of Artificial Neural Networks (ANN) is investigated before and after demonetization in India. Demonetization is the act by government of stripping a currency unit of its status as legal tender. Nine stocks and CNX NIFTY50 index are considered for future value prediction. Nine stocks are subdivided in terms of volatility and capitalization. Dataset for training, testing and validation of each stock under consideration is of at least eight years. Multilayered Neural networks are trained by Levenberg-Marquardt algorithm, hidden layer transfer function is tangent sigmoid, and output layer transfer function is pure linear. Several networks are made by varying the number of neurons to achieve minimum Mean Squared Error (MSE) for an optimum accuracy. Regression values found during training state are 0.999 for all networks that depicts high efficiency of Neural Network designed. Predicted values by the networks designed are validated with actual values before and after demonetization in India.