CHURN RATE: como reduzir em empresas de telecomunicações utilizando aprendizado de máquina

Italo Giullian Carvalho de Albuquerque; Matheus Henrique Bertuci; Bruna Araujo Candeia; Natália de Oliveira Gomes

doi:10.31510/infa.v18i2.1183

Authors

Italo Giullian Carvalho de Albuquerque Universidade Federal do ABC – Santo André – São Paulo – Brasil https://orcid.org/0000-0003-2766-5558
Matheus Henrique Bertuci Faculdade de Tecnologia de Catanduva – Catanduva – São Paulo – Brasil https://orcid.org/0000-0003-4621-8788
Bruna Araujo Candeia Universidade Federal de Pernambuco – Recife – Pernambuco – Brasil https://orcid.org/0000-0002-6747-795X
Natália de Oliveira Gomes Faculdade de Ciências Humanas do Sertão Central – Salgueiro – Pernambuco – Brasil https://orcid.org/0000-0003-0901-6351

DOI:

https://doi.org/10.31510/infa.v18i2.1183

Keywords:

churn rate, customers, data

Abstract

Churn rate corresponds to the cancellation fee related to the use of products/services offered by companies. Telecommunications companies, for example, tend to have high churn rate values due to the high competition in the sector. Therefore, it is essential that companies are able to predict when and why a customer will stop using their services, so that actions can be taken to avoid or at least minimize churn. The premise of this work is to create and compare churn rate predictability models from data from a telecommunications company in order to provide it with information for making more assertive decisions in actions that minimize customer cancellations. Data from a telecommunications operator with a large customer base were used. The database was treated to implement and carry out experiments with machine learning algorithms and data mining techniques. Finally, seven hypotheses were tested which showed that: (H1) the highest turnover rate is in female clients; (H2) old customers tend to have higher turnover; (H3) monthly fees for those who have video streaming are higher; (H4) costs tend to decrease as subscription time passes; (H5) clients who do not have dependents tend to churn more than clients who have dependents; (H6) customers who do not have technical support are more likely to churn; and (H7) customers who have the online backup service tend to churn less.

Downloads

Download data is not yet available.

Metrics

Metrics Loading ...

References

Batista, G. (2003). Pré-processamento de dados em aprendizado de máquina supervisionado. In: Tese (Doutorado) – Curso de Instituto de Ciências Matemáticas e de Computação, São Carlos.

Braga, A. P. et al. Redes neurais artificiais: teoria e aplicações. Rio de Janeiro: LTC, 2000. 250p.

Fonseca, J. (2002). Metodologia da pesquisa científica. Fortaleza: UECE.

Gil, A. (2016). Como elaborar projetos de pesquisa. 4. ed. – São Paulo: Atlas.

Gomes, P. (2019). Conheça as etapas do pré-processamento de dados. Datageeks: https://www.datageeks.com.br/pre-processamento-de-dados/, abril.

Han, J., Kamber, M. (2006). Data Mining: Concepts and Techniques. 2. ed. – São Francisco: Elsevier.

Honda, H., Facure, M., Yaoha, P. (2017). Os três tipos de aprendizagem de máquina. https://lamfo-unb.github.io/2017/07/27/tres-tipos-am/, abril.

Monard, M., Baranauskas, J. (2003). Sistemas Inteligentes Fundamentos e Aplicação. 1. ed. – Barueri: Manole Ltda.

Murteira, B. (1993). Análise Exploratória de Dados: estatística descritiva. McGraw Hill.

Porto, F., Ziviani, (2014) “Seminário de Grandes Desafios da Computação no Brasil” – Rio de Janeiro.

Radosavljevik, D., Putten, P. Van Der, & Larsen, K. (2010). The Impact of Experimental Setup in Prepaid Churn Prediction for Mobile Telecommunications: What to Predict, for Whom and Does the Customer Experience. Trans. MLDM, 3(2), 80–99.

Reis, E. (1997). Estatística multivariada aplicada. Lisboa.

Rossi, R. (2015). Classificação automática de textos por meio de aprendizado de máquina baseado em redes. In: Tese (Doutorado) – Universidade de São Paulo, São Carlos.