Aplicação de um modelo de aprendizagem de máquina para predição de churn: um estudo de caso

Abstract

Machine learning is a computational data analysis algorithm that automates the construction of analytical models. The field of application of these models is quite vast, being used in engineering, medicine, public security, credit operations and several other areas. The present work uses the random forest model to classify customers of a retail company that will no longer use the services provided. The database used was a sample of the total number of customers and is composed of all customers who had their services canceled in the last six months and active customers randomly chosen. The data set used was divided into two other sets, one for training and the other for testing the predictive model. The choice of the random forest model proved to be quite assertive since the algorithm's performance and validation statistics reinforce the predictive strength of the model, presenting an accuracy of 98% and a precision of 97%.