Loading...
Publication
Aplicação de métodos de Inteligência Artificial e Machine-Learning para identificar os fatores condicionantes da subnutrição na região do Sahel
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 7.63 MB | Adobe PDF |
Abstract(s)
A ONU declarou 2016-2025 como a Década de Ação pela Nutrição, e o objetivo 2.2 dos Objetivos de
Desenvolvimento Sustentável (ODS) é "Acabar com todas as formas de malnutrição". O continente
africano, especialmente o Sahel, é fortemente afetado pela subnutrição infantil, onde 40% das crianças
menores de 5 anos sofrem deste problema, com alguns países superando a média continental. Este
trabalho utiliza técnicas de Machine Learning (ML) e Inteligência Artificial (IA), como Random Forest
para regressão e classificação, e Boosted Regression Trees, para identificar os principais fatores que
influenciam a subnutrição infantil.
Os dados utilizados incluem variáveis sociodemográficas dos inquéritos do projeto Demographic and
Health Survey, além de variáveis climáticas e geográficas de fontes pouco exploradas em estudos
anteriores. A análise foi realizada nas regiões administrativas de nível 2, para aproximar os resultados
da escala de tomada de decisão política.
Os resultados mostram que fatores como as condições de vida das mães, a disponibilidade de água e
conflitos armados são determinantes importantes da subnutrição. A modelação espacial com ML
revelou-se eficaz na representação de interações complexas entre fatores socioeconômicos e ambientais.
Em particular, o Random Forest destacou-se pela capacidade de lidar com grandes volumes de dados,
oferecendo resultados robustos para identificar padrões de subnutrição.
Este trabalho sugere que a combinação de variáveis ambientais e socioeconômicas pode melhorar a
análise da subnutrição infantil e contribuir para o desenvolvimento de políticas mais eficazes na região
do Sahel.
The UN declared 2016-2025 as the Decade of Action for Nutrition, and defined objective 2.2 of the Sustainable Development Goals (SDGs) as “End all forms of malnutrition”. As the African continent is one of the most affected by child malnutrition, where 40% of children under 5 years old live with this problem, with some countries in the Sahel standing out for this figure being higher than the average for the continent. This work sought to use Machine learning and Artificial Intelligence techniques, such as Random Forest for regression and classification, and Boosted Regression Trees, to find the main drivers of child malnutrition. Like previous studies, the data source used for the prevalence of malnutrition in the region, among other sociodemographic variables, were the surveys carried out by the Demographic and Health Survey project. In addition to these, climatic and geographic variables from other sources rarely used in previous studies are also used. This analysis was carried out at the scale of level 2 administrative regions in order to approximate the results of the scale of decision-making by political powers. The results show that factors such as mothers' living conditions, water availability and armed conflicts are among the main determinants of malnutrition. Spatial modelling with ML has proven effective in representing complex interactions between socioeconomic and environmental factors. In particular RF models stood out for their ability to handle large amounts of data and providing robust results in identifying patterns of malnutrition. This work offers new perspectives for the study of child malnutrition, suggesting that combining environmental and socioeconomic variables can significantly improve analysis and contribute to the development of more effective policies in the Sahel region.
The UN declared 2016-2025 as the Decade of Action for Nutrition, and defined objective 2.2 of the Sustainable Development Goals (SDGs) as “End all forms of malnutrition”. As the African continent is one of the most affected by child malnutrition, where 40% of children under 5 years old live with this problem, with some countries in the Sahel standing out for this figure being higher than the average for the continent. This work sought to use Machine learning and Artificial Intelligence techniques, such as Random Forest for regression and classification, and Boosted Regression Trees, to find the main drivers of child malnutrition. Like previous studies, the data source used for the prevalence of malnutrition in the region, among other sociodemographic variables, were the surveys carried out by the Demographic and Health Survey project. In addition to these, climatic and geographic variables from other sources rarely used in previous studies are also used. This analysis was carried out at the scale of level 2 administrative regions in order to approximate the results of the scale of decision-making by political powers. The results show that factors such as mothers' living conditions, water availability and armed conflicts are among the main determinants of malnutrition. Spatial modelling with ML has proven effective in representing complex interactions between socioeconomic and environmental factors. In particular RF models stood out for their ability to handle large amounts of data and providing robust results in identifying patterns of malnutrition. This work offers new perspectives for the study of child malnutrition, suggesting that combining environmental and socioeconomic variables can significantly improve analysis and contribute to the development of more effective policies in the Sahel region.
Description
Keywords
Subnutrição Infantil Machine Learning Modelação Espacial Sahel