The popularity of the last article forces us to publish this article this soon. In the last article, we discussed a few performance metrics used for classification problems.
We saw confusion matrix is most commonly used with class output models, however can also be used with probability output models using a threshold probability. We also saw closely linked metrics like KS, Lift and gain. These metrics are generally used when the objective is to target a few out of many. These metrics also help us find out what will be the approx. new response rate if the targeting were revised as per the model.In this article we will take a look at few more metrics of evaluation for classification problems.
Author: Tavish Srivastava