AI models predict adolescent substance use with longitudinal and graph data

Researchers have developed advanced models to predict the onset of substance use in adolescents using data from the ABCD Study. By comparing cross-sectional, longitudinal, and graph-based approaches, they found that longitudinal models consistently outperformed baseline predictions. Temporal XGBoost showed the strongest standalone performance, while Temporal Graph Convolutional Networks (T-GCNs) offered complementary risk signals. Combining these methods through score-level stacking achieved the best results, identifying key predictors such as peer deviance, age, and parental monitoring. AI

IMPACT Demonstrates advanced AI techniques for early identification of adolescent substance use risk, potentially aiding prevention efforts.

RANK_REASON Academic paper detailing a new predictive modeling approach for adolescent substance use. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• ABCD Study
• AUC-ROC
• Hugging Face
• Temporal Graph Convolutional Networks
• T-GCNs
• XGBoost

• paper
• safety
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →