Machine Learning Evaluation

How can we evaluate the success of a machine learning model? For regression, we can simply compute and compare loss functions. For classification however, the loss function is not really intuitive. Instead we introduce simple concepts such as the confusion matrix, and precision/recall. We also talk about the ROC curve, which can help us to choose the right classification parameters.

[jitteryhands]

This was a great explanation! Covered nearly everything I learnt in my class. The confusion matrix is, well, confusing but also quite logical.