DEVELOPMENT OF HYBRID TIME SERIES MODELS FOR FORECASTING AUTUMN RICE USING ARIMAX- ANN AND ARIMAX-SVM.

Time series forecasting is a very active research topic in the domain of science and engineering. The study of forecasting in time series analysis has become a powerful tool in different applications in the agricultural field. Climate change is another major concern in world wide, many researchers are trying to understand its impact on growth and production of crops. Keeping this importance of climate on development and production of crops, an attempt has been made to develop the time series models for forecasting production of Autumn rice in Assam. Yearly data on production of all selected crop and all selected climatic variables have been used for forecasting from the year 1981 to 2018. The data from 1981-2007 were used for model building and 2008 - 2018 were used for checking the forecasting performance of the model. The statistical software viz., SPSS, R; were used for modelling and forecasting of production of agricultural crops in Assam. In this study ARIMAX, ANN, SVM time series models and hybrid of both ARIMAX-ANN, ARIMAX-SVM were used to analyse the past behaviour of production of Autumn rice related with selected climatic variables in order to make inferences about its future behaviour. ARIMA (2,1,2) for production of Autumn rice is applied along with all the weather variables over the growth period of the crop for estimation of production of Autumn rice. The value of MAE under training set for different models ARIMAX (2,1,2), ANN (03:4s:1l), SVM, ARIMAX-ANN and ARIMAX-SVM are found to be 38705.376, 35438.910, 34335.563, 31967.161 & 29484.631 respectively, whereas the value of MAE under testing set are found to be 13502.977, 12430.576, 11149.712, 9436.347 & 9176.531 respectively. Based on these results the model ARIMAX-SVM can be recommended for forecasting of production of crop because of the minimum value of MAE both under training and testing set.