Learning Data Mining with Python - Second Edition

Setting parameters

Almost all parameters that the user can set, letting algorithms focus more on the specific dataset, rather than only being applicable across a small and specific range of problems. Setting these parameters can be quite difficult, as choosing good parameter values is often highly reliant on features of the dataset.

The nearest neighbor algorithm has several parameters, but the most important one is that of the number of nearest neighbors to use when predicting the class of an unseen attribution. In -learn, this parameter is called n_neighbors. In the following figure, we show that when this number is too low, a randomly labeled sample can cause an error. In contrast, when it is too high, the actual nearest neighbors have ...