Clustering Validation Criteria Selection by Meta-Learning
In the Data Mining area, there are different methods that can be adopted for analyzing
large volumes of data. Among them, data clustering stands out. As this task consists
of knowing nothing about the real partition, a careful analysis should be done, and for
that validation criteria exists. Although many criterias have been proposed, it is still a
challenging task indicate which criterion is most suitable for each application scenario.
Considering this fact, the present project proposes the use of meta-learning concepts
for the automatic selection of each relative cluster validation criteria to use, inherently
considering the characteristics of each problem. It is also necessary to identify which are
the meta-attributes for each dataset that influence the performance of the validation
criteria. The results can serve as a guideline for selecting the most suitable index for each
possible application.