Assessing Predictive Models for Tea Yield: A Statistical and Machine Learning Approach in Assam's Biswanath Chariali District
Pal Deka *
Department of Agricultural Statistics, Biswanath College of Agriculture, Assam Agriculture University, Jorhat, India.
Nabajit Tanti
Department of Tea Husbandry and Technology, Biswanath College of Agriculture, Assam Agriculture University, Jorhat, India.
Prasanta Neog
Department of Agro-Meteorology, Biswanath College of Agriculture, Assam Agriculture University, Jorhat, India.
*Author to whom correspondence should be addressed.
Abstract
Climatic factors significantly impact Assam tea production. The tropical climate of Assam, characterized by high precipitation and temperatures up to 36°C during the monsoon, creates ideal conditions for tea cultivation, contributing to the region's unique malty flavor. Here, in this study an attempt has been made to bring a comparison among statistical and machine learning models in prediction of tea production and evaluate an optimal model among them. A time span of last 23 years data were collected from Biswanath College of Agriculture under Assam Agriculture University situated at Biswanath Chariali district. The study has found that mean absolute percentage error of random forest regression model is 6.49 percent followed by decision tree (7.3 percent) and linear regression model (7.5 percent). From the evaluation metrics, random forest algorithm fits well in comparison to decision tree and linear regression. This study could be generalized to comparison among more predictive machine learning models.
Keywords: Assam tea, prediction, machine learning, ; climatic factors