Neto, Miguel de Castro Simões FerreiraJardim, João Bruno Morais de SousaPires, Inês Tomás2024-02-192024-02-192024-01-29http://hdl.handle.net/10362/163771Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business IntelligenceThis study focuses on developing a predictive model for customer churn in a Portuguese bank, using machine learning techniques. Following the CRISP-DM methodology, the analysis encompasses comprehensive EDA, data preparation and visualizations, laying the foundation for model selection. Whitin the subset of evaluated models, such as tree-based and ensembled models, Gradient Boosting emerges as a standout performer, demonstrating notable predictive capabilities. Beyond the identification of customers at risk to churn, this model provides valuable insights, crafting proactive retention strategies. The precision in identifying customers with a high probability of churn enhances informed decision-making. For that reason, an interactive dashboard is developed to empower stakeholders in addressing potential churn risks. These findings underscore the importance of leveraging machine learning in banking scenarios, emphasizing the potential for predictive analytics to enhance customer retention strategies and overall business outcomes.engBanking SectorBusiness IntelligenceCustomer ChurnMachine LearningPower BISDG 8 - Decent work and economic growthSDG 9 - Industry, innovation and infrastructuremaster thesis203524896