4 - Heuristic methods for data clustering

Abstract

Data clustering is a major research area having diverse applications in certain fields, which involves recognition of patterns, machine learning, and mining massive datasets. Also, it is a process of grouping the data, which are similar to each other. As the clustering issue on numerical databases can be modeled as a major combinatorial optimization problem, different research are done on the development of heuristic strategies for determining suboptimal solutions in a certain period of time. However, the majority of the heuristic clustering strategies suffered from the issue of being sensitive to the initialization and does not guarantee to offer qualitative outcomes. In this chapter, different heuristic clustering strategies are devised to provide improved results for clustering massive data. Also, the principles utilized in the heuristic methods are illustrated for getting a brief idea about heuristic methods. Here, the heuristic algorithms are splitted into five types that involve local search versus global search, single-solution versus population-based, hybridization, parallel metaheuristics, and nature-inspired techniques. These methods are briefly elaborated, and the application of the algorithm in data clustering is explained in the chapter. Moreover, algorithms and examples of different algorithms for data clustering are illustrated.