Early Corruption Detection System Using Machine Learning Algorithms

• Autor: Heredia Soto, Maria Fernanda
• Formato: tesis de grado
• Fecha de Publicación: 2025
• Institución: Universidad de Lima
• Repositorio: ULIMA-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorio.ulima.edu.pe:20.500.12724/23707
• Enlace del recurso: https://hdl.handle.net/20.500.12724/23707
• Nivel de acceso: acceso abierto
• Materia: Pendiente; https://purl.org/pe-repo/ocde/ford#2.11.04

This work addresses the development of an early corruption detection system using advanced data analysis and artificial intelligence algorithms. The system uses open data collected from the Anti-Corruption Observatory of Peru’s Comptroller General’s Office. A dataset of 2815 public entities evaluated through the “Índice de Riesgos de la Corrupción e Inconducta Funcional”, with scores ranging from 0 to 100, was structured. The primary task involved classification into six risk levels, from very low to very high. Histogram-based Gradient Boosting (HGB) was applied, achieving an accuracy of 89.6%. Regression tasks on the raw INCO scores and experiments with Large Language Models were also conducted. The system represents an early stage, yet scalable, tool to support public sector transparency. Future work proposes the integration of explainable AI for improved transparency and real-time policy support.