Jesus, Frederico Miguel Campos Cruz Ribeiro deCunha, Soraia Sofia Santiago da2023-02-142023-02-142023-01-25http://hdl.handle.net/10362/149182Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business AnalyticsThe objective of this project is to create a predictive model that will decrease customer churn in a Portuguese bank. That is, we intend to identify customers who could be considering closing their checking accounts. For the bank to be able to take the necessary corrective measures, the model also aims to determine the characteristics of the customers that decided to leave. This model will make use of customer data that the organization already has to hand. Data pre-processing with data cleansing, transformation, and reduction was the initial stage of the analysis. The dataset is imbalanced, meaning that we have a small number of positive outcomes or churners; thus, under-sampling and other approaches were employed to address this issue. The predictive models used are logistic regression, support vector machine, decision trees and artificial neural networks, and for each, parameter tuning was also conducted. In conclusion, regarding the customer churn prediction, the recommended model is a support vector machine with a precision of 0.84 and an AUROC of 0.905. These findings will contribute to the customer lifetime value, helping the bank better understand their customers' behavior and allow them to draw strategies accordingly with the information obtained.engCustomer churn predictionBankingMachine learningSupervised learningmaster thesis203227581