Summary

In this chapter, we have discussed evaluation metrics in Kaggle competitions. First, we explained how an evaluation metric can differ from an objective function. We also remarked on the differences between regression and classification problems. For each type of problem, we analyzed the most common metrics that you can find in a Kaggle competition.

After that, we discussed the metrics that have never previously been seen in a competition and that you won’t likely see again. Finally, we explored and studied different common metrics, giving examples of where they have been used in previous Kaggle competitions. We then proposed a few strategies for optimizing an evaluation metric. In particular, we recommended trying to code your own custom cost functions and provided suggestions on possible useful post-processing steps.

You should now have grasped the role of an evaluation metric in a Kaggle competition. You should also have a strategy to deal with every common...