It would be nice to have some kind of measure around how sure the model is about classifications. In some scenarios, a low confidence (or high neutrality) makes the classification irrelevant and it might be important to know about it.
Maybe there is a better name for what I'm trying to express with this "score" but the general idea is that it answers the question: How confident can I be about the most likely category?
In the scenario where two or more categories are equally probable this "score" should be 0. In the scenario where we have one category with 100% probability and all other categories probabilities are 0% the "score" should be 1. The "score" should range from 0.0 to 1.0 depending on how the probabilities relate to each other.
It would be nice to have some kind of measure around how sure the model is about classifications. In some scenarios, a low confidence (or high neutrality) makes the classification irrelevant and it might be important to know about it.
Maybe there is a better name for what I'm trying to express with this "score" but the general idea is that it answers the question: How confident can I be about the most likely category?
In the scenario where two or more categories are equally probable this "score" should be 0. In the scenario where we have one category with 100% probability and all other categories probabilities are 0% the "score" should be 1. The "score" should range from 0.0 to 1.0 depending on how the probabilities relate to each other.