One class classification is recommended when class distribution are not equally distributed or when minority classes are rare. Then Equal Error Rate (EER) is the suitable parameters if we use Accuracy then model will not be properly evaluated. Now if the classes are balanced then we can use Accuracy with Sensitivity and Speficity. Yes or no?
Single class or multi-class classification is based on the size of data. One class classification is better for small label data. However., when the data is large multi-class classification procedure is preferred.
i think you can evaluate on the metrics you mentioned to measure how accurate is your model. You should also evaluate using F1-score that comprises precision and recall. accuracy with f1-score give you a better insight on the whole model performance.
not to forget that, cross validation with k number of fold gives a better understanding on the metrics measured.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification