Predicting Recurring Telecommunications Customer Support Problems Using Deep Learning

• Vitor Castro ^[1] ; Carlos Pereira ^[2] ; Victor Alves ^[1]
  1. [1] Universidade do Minho

     Universidade do Minho

     Braga (São José de São Lázaro), Portugal

     
  2. [2] NOS Comunicações (Senhora da Hora, Portugal)
• Localización: Intelligent Data Engineering and Automated Learning – IDEAL 2020. 21st International Conference: Guimarães, Portugal; November 4–6, 2020. Proceedings / Cesar Analide (ed. lit.), Paulo Novais (ed. lit.), David Camacho Fernández (ed. lit.), Hujun Yin (ed. lit.), Vol. 2, 2020 (Part II), ISBN 978-3-030-62365-4, págs. 184-193
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • In search of a better quality of experience and more revenue, telecommunication companies are searching for proactive ways of dealing with unsatisfactory user experiences and predicting customer’s behavior. Customer Support (CS) is one of the key areas of customer satisfaction. A good CS enables customers to have a smooth interaction with the company and the services provided when there are doubts or malfunction. Frequently, the problems reported by customers are not resolved in the first interaction, which leads to greater dissatisfaction with the service provider and possibly to future churn. If the company knows in advance of a possible recurrence, it can respond and try to fix the problem without customers noticing or being affected. In this article, a data set of customer data, CS data, and historical service are used to create a deep learning-based model for predicting customer recurrence. Deep neural networks are well-known for their capability to model complex problems when compared to classical machine learning algorithms. The obtained model, with a decision threshold most appropriated for the business needs, presented an F1-score of 60% and AUC-ROC of 61%, with a Recall and Precision of the recurrent class of 29% and 21%, respectively.