A Compendium Probabilistic Prospective for Predicting Coffee Crop Yield Based on Agronomical Factors

Abstract

Coffee has a significant part in the economies of many African, American, and Asian nations since it is a perennial crop and a product used in daily life. For agricultural system modelers, it is still difficult to anticipate coffee output with any degree of accuracy given the environmental, meteorological, and soil fertility factors that influence it. Out of the 103 different varieties of class coffee bean variation that are commercially traded worldwide, the two most significant types of coffee assortment filled in India are Arabica and Robust. In order to investigate and construct a predictive model for the development of coffee planters to make accurate judgments in time during bad conditions in advance, we are taking the most important plantation crop in India, namely, coffee. In order to quantify the effect of agronomic parameters to obtain a decent coffee yield, we thus present a framework for coffee yield prediction utilizing machine learning probabilistic techniques. Here, we take into account the historical dataset from the Central Coffee Research Institute (CCRI), Karnataka, for the year (2008-2019). We are taking into account agronomic elements like Age, Soil Nutrients: Organic Carbon (OC), Phosphorus (P), Potassium (K), Alkaline (pH), and Respective Yield Obtained in Chikkamagaluru Region, Karnataka State, India, for the forecast of coffee yield. Multiple Linear Regression, Lasso Regression, and Elastic Net Regression are three different predictive regression algorithms that are used for prediction; the results of each are compared, examined, and tabulated. The best accurate coffee yield estimate during the autonomous testing phase was produced using an elastic net (ENET) regression model ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}2=0.26$</tex> kg ha 1, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{RMSE}=136.95$</tex> kg ha 1, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{MAE}=111.41$</tex> kg ha 1). This contrasted with the less accurate Lasso regression model ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}=0.25$</tex> kg ha - 1, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{RMSE}=137.64$</tex> kg ha <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−1</sup> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{MAE}=111.96$</tex> kg ha - 1) and MLR ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{R}^{2}=0.25$</tex> kg ha <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−1</sup> , <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{RMSE}=137.71$</tex> kg ha <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−1</sup> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{MAE}= 112.02$</tex> kg ha - 1). In this work, the ENET model replaces the Lasso and MLR models as an enhanced class of artificial intelligence models for smallholder farm coffee yield prediction. This is a distinctive addition to the area of agronomy, particularly in terms of the right selection of the best soil characteristics, coffee variety, and age that may be employed in the forecasting of the ideal coffee production.