Modelo de aprendizaje automático para la predicción de la calidad del café

Abstract

This paper develops a machine learning model (ML) for the qualification of coffee quality in Colombia. With the support of the Colombian coffee quality office, Almacafé, a process of measuring various attributes of green coffee is carried out, over samples from different parts of Colombia, which is used as input data of the model. Then a process of roasting and preparation of coffee drink is carried out to perform the qualification of coffee quality, which is denoted by a score assigned to coffee attributes such as its aroma, its body, cup cleaning, among others (Output data or labels). Subsequently, the chosen machine learning algorithms are trained with the collected data set, implementing a classification approach and a regression approach. The performance measures of these are analyzed and the required adjustments are made. Finally, cross validation of the model is performed.