Table 2. Model selection results.
| Precision | Recall | Fβ = 1 | AUC | |||||
|---|---|---|---|---|---|---|---|---|
| Classifiers | avg | (std) | avg | (std) | avg | (std) | avg | (std) |
| Naïve Bayes | .67 | (.028) | .68 | (.032) | .68 | (.026) | .80 | (.022) |
| SVM Linear | .77 | (.033) | .63 | (.034) | .69 | (.027) | .77 | (.018) |
| SVM RBF | .72 | (.044) | .63 | (.036) | .67 | (.028) | .76 | (.019) |
| Random Forest | .80 | (.035) | .64 | (.037) | .71 | (.030) | .84 | (.018) |
| C4.5 | .79 | (.037) | .62 | (.042) | .70 | (.031) | .80 | (.020) |
Model selection results averaged over 10 runs (10 cross-fold validation per run). Random Forest model proved to be better in terms of precision and AUC measure, whereas Naïve Bayes is better with respect to recall. In terms of Fβ = 1 measure Random Forest and C4.5 are almost equal.