A Comparative Analytical Study of Many Regression Model Approaches, Arima Model and a Hybrid Model for Forecasting Area, Production, and Productivity of Coconut in Kerala, India
Vaisakh Venu
*
Department of BEAS, Kelappaji College of Agricultural Engineering and Technology, KAU, Tavanur-679573, India.
Vipin P. R.
Department of FMPE, Kelappaji College of Agricultural Engineering and Technology, KAU, Tavanur-679573, India.
Prajitha N. K.
Department of BEAS, Kelappaji College of Agricultural Engineering and Technology, KAU, Tavanur-679573, India.
*Author to whom correspondence should be addressed.
Abstract
This study is intended to provide reliable and context-specific forecasting methodologies to support sustainable agricultural planning, resource allocation and policy formulation for the coconut industry in Kerala. It evaluates ARIMA models for Kerala’s coconut production area, production, and productivity. The ARIMA (0, 2, 1) model is preferred for the area of coconut production due to the precision of its residual statistics and the normality of its residual plots. The best fit for production and productivity is provided by the polynomial regression model of orders 3 and 9, respectively, which has lower MAPE, RMSE, and higher R2 values than ARIMA models. The hybrid model, which was created using the best-fitting polynomial and ARIMA models, provides a more accurate representation of the data than either the corresponding polynomial or ARIMA model due to its high R2 and low MAPE values. The hybrid models for area, production, and productivity have MAPE and R2 values of 1.78, 3.56, 3.01, and 0.9799, 0.9664, and 0.9395, respectively.
Keywords: ARIMA, forecasting, mathematical modeling, regression