Adaptive Stock Forecasting Model using Modified Backpropagation Neural Network (MBNN)

Time series based model has been widely applied to estimate the future stock price, and aids investors' decisions and trades. However, due to high rate of volatility and non-linearity of time series, affecting stock market forecasting. To address this, artificial neural network (ANN) and deep neural network (DNN) have been applied in the area of stock price forecasting by various researchers. However, existing model use ANN and DNN with back propagation fails to provide flexible linear or nonlinear relationship among variables and they are difficult to train. The objective of this work is to present a modified back propagation neural network (MBNN) model that can handle huge density of nonlinear data, their relationship and give an optimal strategy for computationally hard problem. Experiments are conducted to evaluate performance of proposed MBNN over existing model in terms RMSE and MAPE. The outcome shows significant performance improvement by MBNN over state-of-art approach.