Predicción de pérdida de clientes bancarios mediante KDD y Aprendizaje Automático

Abstract

Abstract: We find ourselves in a changing world in which companies offering financial services are constantly evolving. There is no shortage of options for customers to choose a bank with better benefits and make the decision to invest their money. Therefore, the main challenges for banking institutions are to retain and avoid customer churn. This research paper proposes the use of the KDD methodology to predict bank customer churn using Machine Learning algorithms. A review of related work is conducted to define variables, best models, and metrics used for this type of problem, through experimentation on a Kaggle customer churn dataset. The results are compared using evaluation metrics (Accuracy, Precision, Recall, F1-Score, and AUC) to show that the best-performing models were Random Forest, XGBoost, and the MLP Neural Network. Additionally, a data visualization prototype is being developed using Dash to analyze customer churn.