Forecasting of Monthly Average Wholesale Prices of Tomato: An Application of Hybrid Time Series Models

The focus of the present study is to give short term forecast of monthly average wholesale prices of tomato using hybrid time series models. For this Hybrid models of the linear seasonal autoregressive moving average (SARIMA) and the nonlinear Artificial Neural Network (ANN) have been considered for estimating and forecasting the monthly average wholesale prices of tomato. The monthly average wholesale prices of tomato from January 2010 to December 2022 have been obtained from different markets of Haryana. The goodness of fit of SARIMA model has been measured using Akaike Information Criteria (AIC), log likelihood (LL), Root Mean Square Error (RMSE) & Mean Absolute Percentage Error (MAPE). The performance of ANN models has been measured using performance measures RMSE & MAPE. The post-sample forecast accuracy has also been measured using MAPE and standard error of prediction (SEP in %). The results of the study showed that the Hybrid (SARIMA + ANN) models have been performed better as compared to SARIMA models for forecasting the monthly average wholesale prices of tomato. The results of this study may be useful for policy makers and various stakeholders in helping them to take appropriate decisions for making arrangements for procurement, distribution, storage, and trading in local and foreign markets, as well as for managing the right inventory in advance.