Mustard price dynamics due to ban on blending: time series intervention model with nonlinear function

Abstract

Intervention analysis is used to study structural changes in data resulting from external events. Traditional time series intervention models, viz. autoregressive in- tegrated moving average model with exogenous variables (ARIMA-X) and artificial neural networks with exoge- nous variables (ANN-X), rely on linear intervention functions such as step or ramp functions, or their combinations. However, when intervention effects are nonlinear, growth models may offer a viable alterna- tive. This study proposed a new algorithm for time series intervention analysis employing ARIMA and ANN models with a nonlinear intervention function. The Hoerl function has been introduced as a nonlinear in- tervention function. To demonstrate the effectiveness of the proposed model, monthly wholesale price data from three markets in Rajasthan, namely, Tonk, Alwar and Sriganganagar during January 2010 to May 2023 have been used. The analysis encompassed a compre- hensive examination across all markets, revealing that the proposed models consistently outperformed the conventional ARIMA-X and ANN-X methodologies in terms of performance and accuracy.