A Comprehensive Review of Evaluation and Fitness Measures for Evolutionary Data Clustering

Data clustering is among the commonly investigated types of unsupervised learning; owing to its ability for capturing the underlying information. Accordingly, data clustering has an increasing interest in various applications involving health, humanities, and industry. Assessing the goodness of clustering has been widely debated across the history of clustering analysis, which led to the emergence of abundant clustering evaluation measures. The aim of clustering evaluation is to quantify the quality of the potential clusters which is often referred to as clustering validation. There are two broad categories of clustering validations; the external and the internal measures. Mainly, they differ by relying on external true-labels of the data or not. This chapter considers the role of evolutionary and swarm intelligence algorithms for data clustering, which showed extreme advantages over the classical clustering algorithms. The main idea of this chapter is to present thoroughly the clustering validation indices that are found in literature, indicating when they were utilized with evolutionary clustering and when used as an objective function.