Questions
1. The definition of "Best class confidence" & "Reference class confidence".
2. What will be the impact of these values being low as 62 to 64 or even in some cases at 37?
3. What is the impact of these values?
Answer
The meaning of these two columns shows the confidence of the class that has been selected as the reference and the confidence of the winning class.
However, there is no way to see what thresholds the classifier has picked up. Therefore, a low confidence value is likely to result in a rejection of the classification.
The main functionality is the following:
- If the confidence of the Best class is less than a certain minimum, then the page will be classified as unknown.
- If the distance between 1 and 2 classes is zero (or a very small number), then the result will still be unknown because the classifier could not make a choice between classes, that are considered identical (in his opinion).
If in this way a harmful class is found that climbs onto other classes, then its confidence must/can be lowered by specifying a value less than 1:
So, the two parameters in question are calculated automatically and for each classifier, counting occurs when training the classifier.
Simply speaking - these columns help to understand why your class was not selected by the classifier.
Comments
0 comments
Please sign in to leave a comment.