Skip to main content


Evaluation is one of the most important aspects of machine learning development. It is the craft of understanding the model and how it works.

Cooks Distance

Cook's distance is used to determine the influence each data point has on the fitted model. In other words, it tells you how much the predictions from your model would change if you excluded certain observations. A large Cook's distance for a particular point means that this point has a substantial influence on the model. This could be because it's an outlier or a highly leveraged point.


Shapely is a framework for evaluating machine learning models.

Confident Learning

Confident learning can if not evaluate the model, then at least it can evaluate the data.



  • The more complex the model, the more difficult it will be to evaluate.