A Study of Effect of Training Period duration and Share Split Event on Predicting Stock Price using Neural Networks
Artificial Neural Network (ANN) is suitable tool for share market prediction due to its ability to learn the data patterns and generalize their knowledge to recognize the future new patterns. In this paper, we introduce an ANN model using Multilayer Feed-Forward NARX network to predict closing price of a share listed in NSE, in particular for SBI. The experiment under study is expected to observe the effect of training period and duration on prediction by ANN. The experiment also studies the performance of ANN before and after share split event. We designed and tested NARX network for SBI data between 2008 and 2016 for different years, to compare the effect of training period on prediction of closing price. We also observed the effect of events like share split on actual closing price and predicted price of SBI.
Keywords - Artificial Neural Network (ANN), Nonlinear Autoregressive with External Input (NARX), stock price prediction, technical analysis, Gradient descent adaptive back-propagation (GDA)