Model performances for classification of presence of any cognitive impairment among different feature sets (MCI and dementia are grouped together). Thresholds balance out the TPR and FPR
| Feature set | Penalty | λ | AUC | Specificity | Sensitivity | PPV | NPV | Number of Features |
|---|---|---|---|---|---|---|---|---|
| NPT | L1 | 10 | 0.929 | 0.836 | 0.927 | 0.81 | 0.938 | 23 |
| MMSE | L1 | 1 | 0.870 | 0.932 | 0.818 | 0.900 | 0.872 | 18 |
| Baseline Linguistic | L2 | 10 | 0.882 | 0.726 | 0.855 | 0.701 | 0.869 | 374 |
| Baseline Linguistic + MMSE | L1 | 1 | 0.913 | 0.863 | 0.873 | 0.828 | 0.900 | 392 |
| Expanded Linguistic | L2 | 100 | 0.883 | 0.767 | 0.873 | 0.738 | 0.889 | 163 |
| Expanded Linguistic + MMSE | L1 | 10 | 0.915 | 0.863 | 0.891 | 0.831 | 0.913 | 181 |
| Expanded Syntactic | L2 | 10 | 0.821 | 0.808 | 0.764 | 0.75 | 0.819 | 31 |
| Expanded Semantic | L2 | 100 | 0.870 | 0.795 | 0.836 | 0.754 | 0.866 | 126 |
| Expanded Lexical | L1 | 10 | 0.844 | 1.000 | 0.618 | 1.000 | 0.777 | 6 |