Modeling and Forecasting of Coconut Area, Production, and Productivity Using a Time Series Model

¹ University of Alkafeel, Najaf, Iraq
² Centurion University of Technology and Management, Odisha, India
³ Jawaharlal Nehru Agricultural University, Madhya Pradesh, India
⁴ VCSG Uttarakhand University of Horticulture Forestry, Uttarakhand, India
⁵ Chaudhary Charan Singh University (Meerut), Uttar Pradesh, India
⁶ University of Delhi, New Delhi, India
⁷ JNKVV College of Agriculture, Rewa, India
⁸ South Ural State University, Chelyabinsk, Russia

^* Corresponding Author: ali.j.r@alkafeel.edu.iq

Abstract

The study aimed to compare ARIMA and Holt's models for predicting coconut metrics in Kerala. The coconut data series was collected from the period 1957 to 2019. Of this, 80% of the data (from 1957 to 2007) is treated as training data, and the rest (20% from 2008 to 2019) is treated as testing data. Ideal models were selected based on lower AIC and BIC values. Their accuracy was evaluated through error estimation on testing data, revealing Holt's exponential, linear, and ARIMA (0,1,0) models as the bestfit choices for predicting coconut area, production, and productivity respectively. After using the testing data, we tried for the forecasting for 2020-2024 using these models, and the DM test confirmed their significant forecasting accuracy. This comprehensive analysis provides valuable insights into effective prediction models for coconut-related metrics, offering a foundation for informed decision-making and future projections.