Aplicação de técnicas de machine learning para classificar a satisfação de clientes de serviços de telefonia celular

Abstract

Measuring the quality of mobile telecommunications services is very important for telephone operators, since there are more smartphones than inhabitants in the Brazilian territory. Quality is sometimes measured through satisfaction, which in turn is collected through questionnaires, so the use of machine learning techniques can help to classify satisfaction, such as understanding the factors of greatest impact. Therefore, the objective of this work is to study and apply machine learning techniques in a database of ANATEL (National Telecommunications Agency), in order to determine which are the most important factors in the telecommunications service for the satisfaction of operators' customers and which algorithm obtains the best results in the classification of satisfaction. The Random Forest, Naïve Bayes, Logistic Regression, K-Nearest Neighbors, Multilayer Perceptron, Support Vector Machine and Decision Tree algorithms were used, and they were evaluated by the Accuracy, Precision, Recall, F1-score and AUC-score metrics. The results of the application showed that Random Forest presented the best results with an accuracy of 0.875 for the test set after cross-validation. Analyzing the importance of the attributes, it is perceived that the quality of the internet, both in stability and speed, is very important in customer satisfaction.