Scholarlink

The growth of Internet connectivity in Brazil has intensified competition among Internet Providers. In this context, customer retention has become a strategic priority, with churn – the proportion of customers who cancel the service – serving as a crucial metric. This study uses churn data and complaints filed with Anatel, Brazil’s telecommunications regulatory agency, by customers of a major Internet company to develop predictive models of churn. By applying natural language processing and machine learning techniques, the study integrates textual data to improve the recall of churn predictions, effectively identifying customers likely to churn. Model interpretability, achieved through Integrated Gradients, identifies words that distinguish customers prone to churn from those who are retained, providing valuable insights for retention strategies.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The Value of Complaints: Churn Prediction in a Major Residential Internet Service Provider Using Textual Data

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wadham Bottacin,
  • Vitor Fontana Zanotelli,
  • Matheus S. De Martin,
  • Pedro de Morais,
  • Rodolfo S. Villaça,
  • Vinícius F. S. Mota,
  • Magnos Martinello,
  • Antonio A. de A. Rocha,
  • Giovanni Comarela

摘要

The growth of Internet connectivity in Brazil has intensified competition among Internet Providers. In this context, customer retention has become a strategic priority, with churn – the proportion of customers who cancel the service – serving as a crucial metric. This study uses churn data and complaints filed with Anatel, Brazil’s telecommunications regulatory agency, by customers of a major Internet company to develop predictive models of churn. By applying natural language processing and machine learning techniques, the study integrates textual data to improve the recall of churn predictions, effectively identifying customers likely to churn. Model interpretability, achieved through Integrated Gradients, identifies words that distinguish customers prone to churn from those who are retained, providing valuable insights for retention strategies.