Exploring Hybrid ARIMA-SVR Approach for Forecast of Gross Cropped Area of Kerala in India

Abstract

The present study aimed at fitting forecasting models to Gross Cropped Area of Kerala State in India using data from 1965-66 to 2022-23. It focused on sole models such as Auto Regressive Integrated Moving Average (ARIMA) and Support Vector Regression (SVR). A hybrid model using combination between Support Vector and time series analysis was also done. The study showed that the most appropriate model based on its Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE) to predict the gross cropped area of the state  is the model of combination between Support Vector and ARIMA models, where it gave more accurate results rather than ARIMA and SVR each separately, that is because the combination between these models combining the flexibility of the time series and the power of Support Vectors, where one of these models compensates the shortage of the other model. Forecasts from ARIMA Model showed a decreasing trend and SVR model showed constancy while Hybrid model showed a decreasing trend in gross cropped area. The study stands itself as a reference for and risk mitigation and effective resource management for agricultural sustainability of the state.