Table 4 Ten-fold cross validation: Comparison of the performance of four machine learning models predicting pre-frailty using all 63 features and selected features (subset of 20 features combination for Fried Frailty Phenotype Classification and 19 features combination for Clinical Frailty Scale Classification)
From: Machine learning models for identifying pre-frailty in community dwelling older adults
Models | AUC (Selected Features) | AUC (All 63 features) | Difference between selected features and all features |
|---|---|---|---|
Fried Frailty Phenotype Classification (not frail: pre-frail) | |||
 Logistic Regression | 0.704 | 0.638 |  + 6.6% |
 Linear Discriminant Analysis | 0.707 | 0.637 |  + 7.0% |
 Support Vector Machine | 0.700 | 0.626 |  + 7.4% |
 Random Forest | 0.722 | 0.701 |  + 2.1% |
Clinical Frailty Scale Classification (not frail: pre-frail) | |||
 Logistic Regression | 0.817 | 0.757 |  + 6.0% |
 Linear Discriminant Analysis | 0.805 | 0.750 |  + 5.5% |
 Support Vector Machine | 0.810 | 0.731 |  + 7.9% |
 Random Forest | 0.800 | 0.776 |  + 2.4% |