I have an EEG dataset for 12 subjects (12 class) which are attached below. I am looking to calculate the accuracy, sensitivity, specificity, F-score for Multi-class.
You can classically calculate classification accuracy by dividing the number correctly classified trials to the total number trials. However, you can not calculate sensitivity, specificity, F-score for Multi-class. I suggest you to use confusion matrix instead of them.
Dear Zaid,
By using Matlab you can calculate them, the following link regarding it
https://au.mathworks.com/matlabcentral/fileexchange/60900-multi-class-confusion-matrix
Regards
Dear Zaid,
You've already added the confusion matrix.
As Dr. Onder writes, the accuracy of classification in multi-class classification is unique. But you can calculate the sensitivity and specificity metrics for each class individually.
For this you need to specify one person as positive, the other as negative class and calculate these values respectively.
In addition,
If you do not want to calculate the sensitivity and specificity separately for each class, you can use the average sensitivity and specificity values (by calculating the average of the calculated sensitivity and specificity values for each class).
Dear Dr. Yasin Kaya, thank you for your answer that exactly what I was doing, but I want to sure that is the correct way to calculate the evaluation measurements for authentication.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification