I am classifying a text document into two categories. I know about the first class but do not know about the second class so I have to set threshold for probability of the first class. If the value of calculated probability for first is less than the threshold then I will assign it to the second class. My question is: how can I set the threshold value if the answer of the calculated probability is very small?
This problem can be solved by re-classifying the learning documents.
You should look for every positively classified documents in each and every class from the learning set. You have to look for the minimum probability required to classify particular document truly in its class. You can use this minimum probability as threshold for the classification model.
- Badges
- Science topic