Publicação
AI models for obesity risk assessment
| Resumo: | Childhood overweight and obesity reflect interconnected behavioural, environmental and social influences. Within the Horizon-Europe PAS GRAS initiative and its longitudinal GicO study, this work investigates whether data-driven methods can reveal meaningful patterns that support detection and prevention at community level. The main objective is to develop Artificial Intelligent models that identify determinants of childhood obesity risk by integrating multi-source information into profiles and indicators that are easily interpretable for early action in real-world settings. Using harmonised datasets covering children’s body composition and physical fitness together with family-context measures,we apply unsupervised learning to characterise population structure and relate profiles to broader behaviours and context. Abrief longitudinal view illustrates how children weight status indicators evolve across assessment moments. Results reveal three clear, interpretable clusters spanning healthier, intermediate and higher-risk profiles. The principal contributions are: (i) an integrated, transparent pipeline from data cleaning to modelling; (ii) visual profile summaries that aid communication with practitioners; and (iii) concise, actionable indicators that connect model outputs to prevention and screening. The materials aim to informmultidisciplinary teamsworking in community programmes and to guide tailored recommendations for specific groups. |
|---|---|
| Autores principais: | Alvarez, Sandra Nathaly Aguilar |
| Assunto: | Inteligência artificial Obesidade infantil Integração de dados Redução de dimensionalidade Clustering Caracterização de perfis Explicabilidade SHAP Percentil de IMC Análise longitudinal |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Instituto Politécnico de Coimbra |
| Idioma: | inglês |
| Origem: | Instituto Politécnico de Coimbra |
| Resumo: | Childhood overweight and obesity reflect interconnected behavioural, environmental and social influences. Within the Horizon-Europe PAS GRAS initiative and its longitudinal GicO study, this work investigates whether data-driven methods can reveal meaningful patterns that support detection and prevention at community level. The main objective is to develop Artificial Intelligent models that identify determinants of childhood obesity risk by integrating multi-source information into profiles and indicators that are easily interpretable for early action in real-world settings. Using harmonised datasets covering children’s body composition and physical fitness together with family-context measures,we apply unsupervised learning to characterise population structure and relate profiles to broader behaviours and context. Abrief longitudinal view illustrates how children weight status indicators evolve across assessment moments. Results reveal three clear, interpretable clusters spanning healthier, intermediate and higher-risk profiles. The principal contributions are: (i) an integrated, transparent pipeline from data cleaning to modelling; (ii) visual profile summaries that aid communication with practitioners; and (iii) concise, actionable indicators that connect model outputs to prevention and screening. The materials aim to informmultidisciplinary teamsworking in community programmes and to guide tailored recommendations for specific groups. |
|---|