Algorithm | Approach 1: 50% comments used for training and 50% for testing | Approach 2: 10-fold cross-validation | |||
---|---|---|---|---|---|
Precision | Sensitivity | F-score | Mean performance score across 10-folds | SD | |
Support vector machines (SVM) | 0.835 | 0.780 | 0.800 | 0.834 | 0.027 |
Random forests | 0.825 | 0.765 | 0.780 | 0.839 | 0.028 |
Decision trees | 0.735 | 0.710 | 0.720 | 0.770 | 0.050 |
Generalised linear models network (GLMNET) | 0.750 | 0.700 | 0.710 | 0.523 | 0.084 |
Bagging | 0.725 | 0.700 | 0.710 | 0.811 | 0.039 |
Maxentropy | 0.670 | 0.670 | 0.670 | 0.014 | 0.009 |
Logitboost | 0.710 | 0.655 | 0.655 | 0.876 | 0.037 |
GLMNET and Bagging have the same F-score, but precision was higher for GLMNET.