Demand forecasting of fresh vegetable product by seasonal ARIMA model

Indian agriculture must remain responsive to managing change and meeting diverse demands of domestic and international stakeholders. Especially when dealing with vegetables with a short shelf life, successful forecasting can be an invaluable way to meet the above mentioned goals. In this paper, we forecast the daily demand for fresh vegetable product (onions) in a Mumbai wholesale market, based on historical data. Of the models developed and tested, a seasonal auto regressive integrated moving average (SARIMA) model outperformed other contenders in terms of forecasting accuracy on both in-sample and two out-of-sample datasets. Results show that the model can be used to forecast with a mean absolute percentage error (MAPE) of 14% which is considered acceptable for products with stochastic demand such as fresh vegetables. In addition to forecasting demand, this paper also aims to provide a practitioners view of ARIMA modelling using Stata that could be used for teaching/learning purposes.