Maize yield forecasting using advanced statistical techniques in India

Abstract

Time series modeling utilizing Autoregressive Integrated Moving Average (ARIMA) and state space (SS) models was established for individual univariate series of maize yield in India. In ARIMA modeling, it is assumed that the underlying parameters remain constant; however, agricultural data is typically collected over time, resulting in time-dependent parameters. The analysis of such data can be conducted through state space procedures utilizing the Kalman filtering technique. This study aimed to assess univariate time series methods for forecasting maize yield in India. The ARIMA (0,1,1) model proved to be suitable; however, the state space model demonstrated superior performance with less error metrics in this empirical study. The models' performances were assessed through a comparison with the observed values.