Comparación e Interpretación de algoritmos de machine learning para la predicción de conversión de clientes potenciales en el Banco de Bogotá

Abstract

This work focuses on the implementation of a predictive model for Banco de Bogotá, aimed at identifying potential customers with a high probability of becoming active clients. A database containing historical information about the bank's users was utilized, allowing for exploratory analysis and the selection of relevant features. Initially, algorithms such as Random Forest, Support Vector Machines (SVM), and logistic regression were applied; however, the results were not satisfactory. To address this, an ensemble model based on SVM was developed, which handles overfitting using Tomek Links, imputes data with K-Nearest Neighbors (KNN), and selects the most important features through logistic regression. As a result, a more robust final model was achieved, enhancing its predictive capacity, as confirmed by sensitivity and specificity metrics.