Autonomous cycle of data analysis tasks for the determination of the coffee productive process for MSMEs

Abstract

Coffee production needs certain levels of efficiency to ensure that the quality of the bean, the roasting process, and in general, the coffee processing methods, achieve financial and environmental sustainability objectives. This requires tasks of monitoring and analyzing of features of the coffee bean, and the roasting process, among other aspects, so that stakeholders of the agro-industrial sector of MSMEs can know what happens in the coffee production and can make better decisions to improve it. In a previous article, three autonomous cycles of data analysis tasks are proposed for the automation of the production chains of the MSMEs. This work aims to instantiate the autonomous cycle responsible for identifying the type of input to transform in the production process, in the case of coffee production. This cycle analyzes the inputs of the production chain (quantity, quality, seasonality, durability, cost, etc.), based on information from the organization and the context, to establish the production process to be carried out. This autonomous cycle is instanced in the coffee production to identify the type of input to transform (bean quality), and to determine the transformation process (level of decrease of the bean during the roasting process and coffee processing method). The quality model is defined by the K-means technique with a performance in the Silhouette Index of 0.85, the predictive model of the level of decrease of beans in the roasting process is defined by Random Forest with a performance in the accuracy of 0.81, and finally, the identification model of the "production method" is carried out by the Logistic Regression technique with a quality performance in the accuracy of 0.72. Among the most important findings is that the autonomous cycle of data analysis tasks based on machine learning techniques is capable of studying the contextual data of coffee production to identify the type of input to be transformed and the coffee transformation process. Another important finding is that the autonomous cycle allows the automation of the production process, leading to improved times and coffee processing.