Publicação

Predicting customer churn in telecommunications: A machine learning approach using tensor representations

Ver documento

Detalhes bibliográficos
Resumo:Customer churn has become an increasingly important concern in the telecommunications sector, where high operational costs and market competition emphasize the need to retain customers. This study aimed to evaluate if advanced methods can improve the identification of customers at risk of churning. Conventional machine learning models revealed limitations in recognizing churners, mainly due to class imbalance, despite achieving a good overall accuracy. To address these challenges, a new method leveraging covariance structures is presented, capturing complex relationships and patterns. In addition, the use of sentiment analysis on customer interactions was explored, highlighting its potential for early churn prediction. While churn prediction remains a difficult task influenced by evolving customer behaviours, this analysis demonstrated that supervised machine learning along with proper feature engineering can enhance churn detection, enabling telecommunication companies to act proactively to improve retention, reduce customer losses and maintain income stability. This work contributes to the adoption of data-driven strategies in customer management, supporting telecommunications companies in addressing churn challenges with greater efficiency.
Autores principais:Barbosa, Duarte Pereira
Assunto:Churn Prediction Telecommunications Imbalanced Data Powers of the Covariance Matrices Sentiment Analysis SDG 8 - Decent work and economic growth SDG 9 - Industry, innovation and infrastructure SDG 10 - Reduced inequalities SDG 11 - Sustainable cities and communities
Ano:2025
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:Customer churn has become an increasingly important concern in the telecommunications sector, where high operational costs and market competition emphasize the need to retain customers. This study aimed to evaluate if advanced methods can improve the identification of customers at risk of churning. Conventional machine learning models revealed limitations in recognizing churners, mainly due to class imbalance, despite achieving a good overall accuracy. To address these challenges, a new method leveraging covariance structures is presented, capturing complex relationships and patterns. In addition, the use of sentiment analysis on customer interactions was explored, highlighting its potential for early churn prediction. While churn prediction remains a difficult task influenced by evolving customer behaviours, this analysis demonstrated that supervised machine learning along with proper feature engineering can enhance churn detection, enabling telecommunication companies to act proactively to improve retention, reduce customer losses and maintain income stability. This work contributes to the adoption of data-driven strategies in customer management, supporting telecommunications companies in addressing churn challenges with greater efficiency.