Modelo de clasificación en Machine Learning para productos del Fondo Nacional de Garantías S.A. - FNG

Abstract

This work details the implementation of a classification model in Machine Learning in order to analyze and evaluate guarantees in the scope of the National Guarantee Fund. Its main objective is to improve efficiency and precision in the detection of inconsistencies in guarantee claims by financial intermediaries, thus optimizing the process. The narrative is based on the author's experience during an internship at the National Guarantee Fund, specifically in areas such as the Vice Presidency of Operations and the Guarantees Subdirectorate, playing the role of technical analyst. The central focus of the work is to identify the financial intermediaries that tend to present the most inconsistencies and define the products and types of processes susceptible to notification. A database that includes all guarantees with some inconsistency at the time of the claim is used to illustrate the implementation of the classification model in Machine Learning. The code presented uses Python and libraries such as pandas, scikit-learn and openpyxl to load, process and build the model. classification. Mathematical and statistical knowledge is applied, such as linear algebra, probability and Boolean logic. In addition, machine learning is implemented using a RandomForestClassifier model, based on decision trees and graph theory, with optimization in model training. The theoretical framework focuses on the selection and preparation of data, the detailed exploration and appropriate choice of classes, until the implementation and evaluation of the model. The suitability of classification algorithms to optimize the prediction of inconsistencies and behaviors of the characteristics of the guarantees. This work contributes to the field of application of Machine Learning in financial entities, offering a specific and practical case study to improve processes in the context of the National Guarantee Fund. It is proposed to examine the implementation of classification models in a specific business context, evaluating their impact on decision making and improving operational efficiency.