Classification report

<aside> 💡 For intuitive understanding of all important metrics → Confusion Matrix & more

</aside>

You can easily write custom metric functions for scikit-learn, but you need to take an extra step if you want to use those metrics in a hyperparameter search. To add a number there, you need to use a scorer instead. → Custom Metric for evaluation

• Metric vs. loss: You actually care about the metric but in training you optimize for loss. The difference is explained in the fastai tutorial. It is much better to calculate the gradient from the loss function which is smooth and has usually no sharp edges. The metric is the thing we actually care about, the loss the thing that is similar to what we care about that has a nicely behaved gradient.

<aside> 💡

Choose the right metric and plot it, both done best with skore !!!

</aside>

Accuracy

More details here: Accuracy

Accuracy is one of the easiest ways to evaluate the performance of your model. But it doesn’T work well for unbalanced data.

<aside> 💡

How often am I right about my prediction?

</aside>