Predictive model for cocoa yield in Santander using Supervised Machine Learning

Supervised Machine Learning represent a good alternative for the agriculture, in the way that it allows to support farmers, government and other stakeholders in the decision-making process based on crop yield forecast, which are defined as the volume of product harvested per unit area. This investigation has as object of study an experimental culture of cocoa in Santander, located in the research center La Suiza, and its purpose is to predict the yield of the crop through a set of photosynthetic, morphological, climatic, chemical and physical variables. Using the Generalized Linear Model (GLM) and the Vector Support Machines (SVM), the explanatory variables with the greatest impact were identified both negatively and positively on the cocoa crop yield variable. The construction and comparison of the results of the two models, was useful to ratify that the explanatory variables: Diameter of the trunk, Phosphorus (P), Magnesium (Mg), % Sand, % Hum/Grav, Radiation, Temperature, Humidity and Rains accumulated are the variables that explain to a greater extent the performance of the cocoa crop.