Frailty classification using machine learning on behavioral and physical health data

Publication

Chao Bian, S. Khan, A. Mihailidis, 2021, “Frailty Classification Using Behavioral and Physical Health Data,” International Journal of Medical Informatics (In Review).

Datasets

• The Survey of Health, Ageing and Retirement in Europe (SHARE) dataset
• 27K samples after data cleaning
• Mixed data types and missing values

Machine Learning Classifiers

• 5 tree-based classifiers
• 3 fold cross validation

Features

22 features in the domains of demographics, walking speed, physical activity level, mobility, activities of daily livings.

Model Performance

Feature importance