Implementación de una metodología basado en IA para el análisis de la gestión pública: un enfoque innovador para combatir la pobreza en la ciudad de San José de Cúcuta.
No hay miniatura disponible
Fecha
2025
Autores
Rodríguez Guevara, Jorge Enrique
Título de la revista
ISSN de la revista
Título del volumen
Editor
Ediciones Universidad Simón Bolívar
Facultad de Ingenierías
Facultad de Ingenierías
Resumen
El proyecto propone una metodología innovadora basada en inteligencia artificial (IA) para
analizar los comportamientos socioeconómicos de la población de San José de Cúcuta, con el fin
de fortalecer la toma de decisiones públicas orientadas a la reducción de la pobreza extrema. La
ciudad enfrenta un escenario complejo marcado por migración masiva, violencia territorial,
desigualdad estructural y limitaciones institucionales que han profundizado las brechas
socioeconómicas. Las metodologías tradicionales de análisis, centradas principalmente en
indicadores monetarios o categorizaciones simplificadas, han demostrado insuficiencia para
comprender la diversidad y la multidimensionalidad de la pobreza en territorios fronterizos. Por
ello, el estudio plantea la necesidad de incorporar técnicas avanzadas de IA como una alternativa
estratégica, rigurosa y replicable para segmentar a la población e identificar con mayor precisión
los núcleos de vulnerabilidad.
El marco conceptual se fundamenta en la pobreza multidimensional, particularmente en la
metodología Alkire-Foster, en los enfoques de desarrollo humano de Amartya Sen y en modelos
contemporáneos de análisis causal. La investigación reconoce la pobreza como un fenómeno
estructural que abarca múltiples dimensiones: educación, salud, condiciones de vivienda, empleo,
acceso a servicios básicos y participación social. Desde la perspectiva tecnológica, se integran
principios de machine learning, aprendizaje no supervisado, inteligencia artificial explicable (XAI)
e interoperabilidad de datos, relevando su utilidad para la gestión pública cuando se aplican de
forma ética, transparente y con criterios robustos de calidad de datos.
Metodológicamente, la investigación se desarrolló en tres fases principales. La primera consistió
en la recolección, depuración y normalización de datos provenientes de fuentes oficiales, censos,
registros administrativos y encuestas. Se aplicaron procesos de limpieza, eliminación de
duplicados, imputación de valores faltantes y codificación categórica, generando un dataset
confiable y representativo. La segunda fase correspondió al análisis exploratorio y segmentación
mediante algoritmos de aprendizaje no supervisado. Se emplearon K-Means++, DBSCAN y
métricas como el coeficiente de Silhouette, correlaciones de Pearson y validaciones piloto para
asegurar la consistencia de los hallazgos. La tercera fase integró los resultados del clustering con
análisis cualitativo a través de un grupo focal, logrando triangulación metodológica y una
interpretación más profunda de los patrones detectados,.
The project proposes an innovative methodology based on artificial intelligence (AI) to analyze the socioeconomic behaviors of the population of San José de Cúcuta, with the aim of strengthening public decision-making focused on reducing extreme poverty. The city faces a complex scenario marked by mass migration, territorial violence, structural inequality, and institutional limitations that have deepened socioeconomic gaps. Traditional analytical methodologies, primarily focused on monetary indicators or simplified categorizations, have proven insufficient to understand the diversity and multidimensionality of poverty in border regions. Therefore, the study argues for the need to incorporate advanced AI techniques as a strategic, rigorous, and replicable alternative for segmenting the population and more accurately identifying areas of vulnerability. The conceptual framework is based on multidimensional poverty, particularly the Alkire-Foster methodology, Amartya Sen's human development approaches, and contemporary models of causal analysis. The research recognizes poverty as a structural phenomenon encompassing multiple dimensions: education, health, housing conditions, employment, access to basic services, and social participation. From a technological perspective, it integrates principles of machine learning, unsupervised learning, explainable artificial intelligence (XAI), and data interoperability, highlighting their usefulness for public administration when applied ethically, transparently, and with robust data quality criteria. Methodologically, the research was developed in three main phases. The first consisted of collecting, cleaning, and normalizing data from official sources, censuses, administrative records, and surveys. Processes such as cleaning, duplicate removal, imputation of missing values, and categorical coding were applied, generating a reliable and representative dataset. The second phase involved exploratory analysis and segmentation using unsupervised learning algorithms. KMeans++, DBSCAN, and metrics such as the Silhouette coefficient, Pearson correlations, and pilot validations were used to ensure the consistency of the findings. The third phase integrated the clustering results with qualitative analysis through a focus group, achieving methodological triangulation and a deeper interpretation of the detected patterns.
The project proposes an innovative methodology based on artificial intelligence (AI) to analyze the socioeconomic behaviors of the population of San José de Cúcuta, with the aim of strengthening public decision-making focused on reducing extreme poverty. The city faces a complex scenario marked by mass migration, territorial violence, structural inequality, and institutional limitations that have deepened socioeconomic gaps. Traditional analytical methodologies, primarily focused on monetary indicators or simplified categorizations, have proven insufficient to understand the diversity and multidimensionality of poverty in border regions. Therefore, the study argues for the need to incorporate advanced AI techniques as a strategic, rigorous, and replicable alternative for segmenting the population and more accurately identifying areas of vulnerability. The conceptual framework is based on multidimensional poverty, particularly the Alkire-Foster methodology, Amartya Sen's human development approaches, and contemporary models of causal analysis. The research recognizes poverty as a structural phenomenon encompassing multiple dimensions: education, health, housing conditions, employment, access to basic services, and social participation. From a technological perspective, it integrates principles of machine learning, unsupervised learning, explainable artificial intelligence (XAI), and data interoperability, highlighting their usefulness for public administration when applied ethically, transparently, and with robust data quality criteria. Methodologically, the research was developed in three main phases. The first consisted of collecting, cleaning, and normalizing data from official sources, censuses, administrative records, and surveys. Processes such as cleaning, duplicate removal, imputation of missing values, and categorical coding were applied, generating a reliable and representative dataset. The second phase involved exploratory analysis and segmentation using unsupervised learning algorithms. KMeans++, DBSCAN, and metrics such as the Silhouette coefficient, Pearson correlations, and pilot validations were used to ensure the consistency of the findings. The third phase integrated the clustering results with qualitative analysis through a focus group, achieving methodological triangulation and a deeper interpretation of the detected patterns.
Descripción
Palabras clave
Inteligencia artificial, Pobreza multidimensional, Gestión Pública, Segmentación socioeconómica, Machine learning
