A comparison of ARIMA & NNAR models for production of wheat in the state of Andhra Pradesh

Abstract

If the data is linear and non-stationary, the models viz. Auto-Regressive (AR), Moving Average (MA), and Auto-Regressive Moving Average (ARMA) models cannot be used. So, another important forecasting technique called Auto-Regressive Integrated Moving Average (ARIMA) with (p, d, q) terms can be used. The best feature of Artificial Neural Networks when it is applied to forecasting data is its inherent capability of nonlinear modeling without any presumption about the statistical distribution of the given data. Model selection criteria based on RMSE for ARIMA and Neural Network Autoregressive (NNAR) models are computed. An appropriate model has to be framed effectively for the production of Wheat data in the state of Andhra Pradesh taken during the period from 1982 to 2022 (for 40 years).