Modelling and forecasting sorghum (Sorghum bicolor) production in India using hierarchical time-series models

摘要

Hierarchical time-series comprises of several dataset maintaining certain hierarchical relationship among them. There are certain specialized strategies, viz. top-down, bottom-up, middle-out and optimal approaches which take care of predicting future values for such multi-level data. For forecasting of individual series at different levels of hierarchy, a method of aggregation or disaggregation is followed. In the present study, these methodologies are investigated thoroughly. Further, state-wise seasonal sorghum [Sorghum bicolor (L.) moench] production data of India is analyzed by employing the hierarchical forecasting approaches. A comparative study on performance of different methods is carried out from the viewpoint of multi-step-ahead forecasts on the basis of Mean absolute error (MAE) and Root mean square error (RMSE). The findings show that the middle-out technique outperforms other approaches and traditional method of forecasting as well. This fact has been confirmed statistically by using pair-wise t-test. Finally, using the middle-out approach, forecasts of sorghum production for 2015 till 2017 have been carried out at all hierarchical levels. For statistical analysis, R software package has been employed.