Evaluation
The evaluation phase is the one where we look for a validation of the results coming from our modeling activities. This phase can be split into two main questions:
- Is the model performing adequately?
- Is the model answering the questions originally posed?
The first of the two questions involves the identification of a proper set of metrics to establish if the model developed possesses the desired properties.
Following previously-shown model families, we are going to show you here how to overcome the following problems:
- Clustering evaluation
- Classification evaluation
- Regression evaluation
- Anomaly detection evaluation
Clustering evaluation
It is quite easy to understand how to evaluate the effectiveness of a clustering model. Since the objective of a clustering model is to divide a population into a given number of similar elements, evaluation of these kinds of models necessarily goes through the definition of some kind of an ideal clustering, even if defined by human judgment. Evaluating...
