Analysis of Investors' Prediction Potential Using Holt Winters Model and Artificial Neural Networks

This study analyses the predictive power of six categories of investors who traded Nifty Index future contracts on the National Stock Exchange of India (NSE). Quality data were collected from The Securities and Exchange Board of India (SEBI). Investors' predictive potential was estimated by analysing the effect of open interest and volume traded on volatility for each category. The Holt-Winters (HW) exponential smoothing model successfully captured seasonality and provided a satisfactory analytical model for linear forecast. Nonlinearity was captured by Artificial Neural Networks (ANN). A resilient backpropagation algorithm with backtracking was used to determine the weights triggered by a logistic activation function for smoothing neurons. The multilayer perceptron network was further trained for time series data on volatility considering volume and open interest as input neurons. Predictive powers were considered best for the method with the least Root Mean Square Error (RMSE). The results suggest that nonlinearity in the data was well captured by the ANN as the RMSE for the ANN was smaller compared to the RMSE for the HW model. The RMSE using ANN was least for Foreign Institutional Investors (FIIs) that suggests that FIIs have better prediction potential compared to other investors.