Table 7.
The classification algorithms performance against subsets of the ADAS-Cog13 items.
| Subset | Algorithm | Without Using Any Demographics | When Using Few Demographics | ||||
|---|---|---|---|---|---|---|---|
| Accuracy % | Sensitivity % | Specificity % | Accuracy % | Sensitivity % | Specificity % | ||
| ADAS-subset1 (baseline) | BayesNet | 79.27% | 84.30% | 74.30% | 88.51% | 89.60% | 87.50% |
| Logistic Regression | 81.83% | 86.10% | 77.70% | 91.01% | 93.60% | 88.50% | |
| C4.5 | 88.59% | 89.00% | 88.20% | 91.48% | 91.60% | 91.40% | |
| ADAS-subset2 | BayesNet | 73.81% | 77.50% | 70.20% | 87.87% | 90.40% | 85.30% |
| Logistic Regression | 77.07% | 81.90% | 72.40% | 90.26% | 93.30% | 87.30% | |
| C4.5 | 86.68% | 86.90% | 86.40% | 91.75% | 91.50% | 92.00% | |
| ADAS-subset3 | BayesNet | 69.80% | 78.10% | 61.60% | 78.80% | 84.90% | 72.80% |
| Logistic Regression | 74.21% | 78.80% | 69.70% | 80.05% | 83.50% | 76.60% | |
| C4.5 | 82.90% | 87.60% | 78.20% | 88.69% | 88.80% | 88.60% | |
| ADAS-subset4 | BayesNet | 74.36% | 78.20% | 70.50% | 87.37% | 89.10% | 85.60% |
| Logistic Regression | 79.40% | 84.10% | 74.80% | 90.35% | 93.10% | 87.60% | |
| C4.5 | 86.19% | 88.80% | 83.60% | 91.94% | 92.60% | 91.20% | |