Table 3.
Confusion matrix for predicting PPA diagnosis for 34 participants using random forests classification.
| Predicted diagnosis | ||||||
|---|---|---|---|---|---|---|
| lvPPA, n (%) | nfvPPA, n (%) | svPPA, n (%) | Controls, n (%) | Accuracy | ||
| Actual diagnosis | lvPPA, n (%) | 8 (89) | 1 (11) | 0 (0) | 0 (0) | 0.924 |
| nfvPPA, n (%) | 0 (0) | 7 (100) | 0 (0) | 0 (0) | 0.981 | |
| svPPA, n (%) | 1 (14) | 0 | 6 (86) | 0 (0) | 0.928 | |
| Controls, n (%) | 0 (0) | 0 (0) | 0 (0) | 11 (100) | 1.000 | |
The overall balanced accuracy of the model was 0.958. True positives in bold type. lvPPA, logopenic variant PPA; nfvPPA, non-fluent variant PPA; svPPA, semantic variant PPA.