An Improved ARIMAX model using ANN and NLSVR for Forecasting Jute production in Assam

Borsha Neog^1*Horindra Gogoi²Mousumi Phukon³Bhabesh Bhagawati⁴Tonmoy Neog⁵

• ¹Department of Agricultural Statistics, Assam Agricultural University, Jorhat, India
• ²Department of Agricultural Economics, Assam Agricultural University, Jorhat, India
• ³Department of Entomology, Assam Agricultural University, Jorhat, India
• ⁴Department of Nematology, Assam Agricultural University, Jorhat, India
• ⁵Department of Geophysics, Banaras Hindu University, Varanasi, India

Word count: 189 Jute, it is often referred as the "golden fibre" and it plays a pivotal role in the agrarian economy of India, particularly in Assam. Despite its ecological advantages and economic potentiality; jute production in Assam faces many challenges due to climatic variability, traditional farming practices and inconsistent forecasting techniques. This research proposed a novel hybrid model integrating ARIMAX-ANN, ARIMAX-NLSVR to enhance the accuracy of forecasting of jute production in Assam. Historical data were taken from directorate of jute development, ministry of Agriculture, Govt. of India and IMD, Guwahati from 1981 to 2018. The performance of the hybrid models was assessed using standard forecasting accuracy metrics including mean absolute error (MAE) and mean absolute percentage error (MAPE). ARIMA (0,1,1) model has been applied with maximum temperature over the growth period of the crop for estimation of production of Jute.The value of MAE and MAPE under training and testing set for different models ARIMAX (0,1,1), ANN (03:4s:1l), NLSVR, ARIMAX-ANN and ARIMAX-NLSVR have been checked. The findings indicate that the hybrid models outperformed better as compared to individual models. These results offering a reliable decision support tool for agricultural planning and management.