TABLE 3.
Classification performance metrics on the best classifier and predictor sets.
| Symptom | Best classifier | AUC a | F‐score | Number of features |
|---|---|---|---|---|
| Global rating of alogia | Linear SVM | 1 | 0.82 | 29 |
| Illogicality | Decision tree | 1 | 0.68 | 2 |
| Poverty of speech | Neural net | 1 | 0.65 | 60 |
| Global rating of hallucinations | LDA | 0.995 | 0.64 | 49 |
| Global rating of avolition/apathy | Gaussian process | 1 | 0.62 | 32 |
| Social inattentiveness | Nearest neighbors | 0.86 | 0.60 | 41 |
| TLC | Naive Bayes | 0.93 | 0.58 | 3 |
| Global rating of bizarre behavior | Linear SVM | 0.93 | 0.55 | 26 |
| Global rating of affective flattening | LDA | 0.995 | 0.54 | 42 |
| Global rating of delusions | Gaussian process | 1 | 0.53 | 46 |
a
AUC was calculated using the train data.