Utilização de classificadores múltiplos com modelo de regressão em sobrevivência, na classificação de clientes

Abstract

Nowadays, in customer relationships, companies are interested in identifying the individuals most likely to abandon a service, in a process identified as churn, to reduce losses in possible interruptions or disconnections of these customers. Several studies have therefore been presented with the aim of predicting this event using multiple classifiers, applied with machine learning techniques. In this work we intend to evaluate multiple classification methods for classifying individuals according to their personal characteristics, using regression models in survival analysis. Initially, the bagging and boosting classifiers were applied and the results obtained were then analyzed.