Table 5. Results of Linear SVM and Random Forest Models.
|
aMCI versus Deficit SCZ |
Linear SVM | Random Forest | ||||
| Validation accuracy (%) | Top-features | Feature weights | Validation accuracy (%) | Top-features | Feature importances | |
| 7 CERAD | 81.0 (13.0) | BNT | -0.6682 | 78.0 (10.7) | BNT | 0.2991 |
| WL | -0.3866 | WLM | 0.2489 | |||
| False Recall | 0.2759 | VFT | 0.1390 | |||
| WLM | ||||||
| 7 CERAD | 97.0 (4.6) | BNT | -0.2699 | 98.0 (4.0) | BNT | 0.1977 |
| - Age | WLM | 0.2606 | WLM | 0.0925 | ||
| - Sex | VFT | 0.2465 | WL | 0.0406 | ||
| - Education | False Recall | |||||
| aMCI versus Nondeficit SCZ | Validation accuracy (%) | Top-features | Feature weights | Validation accuracy (%) | Top-features | Feature importances |
| CERAD | 73.6 (10.9) | BNT | -0.5626 | 71.4 (14.3) | BNT | 0.3410 |
| WLM | 0.3562 | WLM | 0.2226 | |||
| WLRec total | -0.2955 | VFT | 0.1882 | |||
| 7 CERAD, | 100.0 (0.0) | BNT | -0.3144 | 98.9 (3.3) | BNT | 0.1545 |
| - Age | WL | 0.3013 | WL | 0.0573 | ||
| - Sex | False Recall | False Recall | ||||
| - Education | WLM | 0.2719 | VFT | 0.0477 | ||
10-fold cross-validation performances and predictive weights in Support Vector Machine with linear kernel (Linear SVM) and importances in Random Forest Models performed using 7 Consortium to Establish a Registry for Alzheimer’s disease (CERAD) test results as explanatory variables and amnestic mild cognitive impairment (aMCI) versus (non)deficit schizophrenia (SCZ) as dichotomous classes. VFT: Verbal Fluency Test. BNT: Boston Naming Test. WLM: Word List Memory. WL False Recall: Word List Recall, Delayed, false recall. WLRec total: Word List Recognition total.