Modelos de machine learning aplicados a dados de churn bancario

Abstract

In this study, the application of machine learning algorithms was investigated using a synthetic dataset that simulates the churn phenomenon, defined as the closure of a customer’s account, in a banking institution. A cross-validation process was conducted on logistic regression and random forest algorithms, revealing that the latter demonstrated superior performance. Subsequently, the random forest model was evaluated on the test set, yielding recall of 0.45, precision of 0.77, and an F1 score of 0.57. While these results are deemed acceptable, a business-oriented analysis indicates potential for enhancing model performance by reducing the false negative rate.