GULP-KG: Graphical Understanding of Lifestyle Patterns with Knowledge Graphs

Summary: We propose GULP-KG, a knowledge-graph framework that fuses multimodal health data to discover lifestyle patterns and support interpretable reasoning.

Abstract

GULP-KG models lifestyle through typed entities (behaviors, contexts, biomarkers) and relations learned from multi-source sensing and logs. Graph construction aligns heterogeneous streams; graph representation learning enables pattern discovery and counterfactual queries. On benchmarked cohorts, GULP-KG captures recurring lifestyle motifs, supports interpretable retrieval, and improves downstream prediction over feature-only baselines, demonstrating the utility of KG-centered health analytics.