ABSTRACT
This chapter describes components of the overall process of finding clusters from static sets of objects that are represented by numerical data. Models provide mathematical properties that define computer-point-of-view (CPOV) clusters in data. Algorithms are methods that find the CPOV clusters in static data by optimizing a model. There are four types of clustering algorithms corresponding to the four types of partitions but, again, only three distinct mathematical structures. The fuzzy/probabilistic partitions share a common mathematical structure, but the assumptions about data underlying fuzzy and probabilistic clustering models are very different. CPOV cluster analysis comprises computational models and algorithms that search for groups (c-partitions) in unlabeled data. The relationship between physical labels for subsets of labeled data to CPOV labels assigned it by clustering algorithms is discussed in detail in the sequel.
