A new research paper explores the use of in-context learning with the TabPFN foundation model to assess how built environment features impact perceived neighborhood walkability among older adults with mobility impairments. The study found that TabPFN outperformed traditional models like Random Forest and XGBoost on a small dataset, achieving a macro F1 score of 54.89%. Interpretive analysis using Shapley Interaction Quantification (SHAP-IQ) revealed that higher-order feature interactions, such as street circuity combined with drivable roads, were key predictors of perceived walkability. AI
IMPACT Demonstrates the potential of foundation models for specialized, small-dataset tasks and provides advanced interpretability for urban planning insights.
RANK_REASON Research paper detailing a novel application of an existing model (TabPFN) and interpretation technique (SHAP-IQ) to a specific domain (urban planning/mobility). [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- Neighborhood Environment Walkability Scale: validity and development of a short form
- news article
- random forest
- SHAP-IQ
- Shapley Interaction Quantification
- TabPFN
- XGBoost
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
