A hybrid wavelet based neural networks model for predicting monthly WPI of pulses in India

Abstract

The high prices of pulses continue to be the pain point for both consumers and policymakers. In India, the wholesale price index (WPI) is the main measure of inflation. WPI measures the price of a representative basket of wholesale goods. Therefore, accurate forecasting of WPI is necessary by using some advanced statistical techniques. In the present investigation, Wavelet and artificial neural network (Wavelet-ANN) hybrid models are used for multi-step-ahead forecasting of monthly WPI of pulses. The original series is decomposed into the low frequency and high frequency components using Maximal Overlap Discrete Wavelet Transform (MODWT) based on Haar wavelet filter. Subsequently, suitable artificial neural network (ANN) model was fitted to decomposed series before they are combined and predicted using Inverse Wavelet Transform (IWT). A comparative assessment of hybrid models as well as individual counterpart revealed that the hybrid models give significantly better results than the classical artificial neural network (ANN) model for all tested situations.