MÉTODOS DE MACHINE LEARNING PARA CLASSIFICAÇÃO DA TEMPERATURA NO PROCESSO DE MISTURA EM UMA PLANTA INDUSTRIAL

Abstract

The goal of this paper is the application and analysis of Machine Learning methods to classify the temperature resulting from the liquids mixing process in a didactic industrial plant located in the UTFPR of the Cornélio Procópio city. The applicable methods for this classification are k-nearest neighbors (KNN), Decision Trees, Random Forest and Naive- Bayes. For this task, the input variables considered are the percentage of valve opening, the flow and the opening time; and the output variable is temperature, discretized into 5 classes. The performance of the algorithms is analyzed considering the accuracy and relevant statistical measures and the implementations of the methods are made using the R software. The results show a good performance of the algorithms in the task of classifying the temperature of the liquid mixing process, with accuracy above 90%.