Nature Inspired Clustering - Use Cases of Krill Herd Algorithm and Flower Pollination Algorithm

Piotr A. Kowalski, Szymon Lukasik, Malgorzata Charytanowicz, Piotr Kulczycki

Abstract: Nature inspired metaheuristics were found to be applicable in deriving best solutions for several optimization tasks, and clustering represents a typical problem which can be successfully tackled with these methods. This paper investigates certain techniques of cluster analysis based on two recent heuristic algorithms mimicking natural processes: the Krill Herd Algorithm (KHA) and the Flower Pollination Algorithm (FPA). Beyond presenting both procedures and their implementation for clustering, a comparison with regard to quality of result was performed for fifteen data sets mainly drawn from the UCI Machine Learning Repository. As a validation of the clustering solution, the Calinski-Harabasz Index was also applied. Moreover, the performance of the investigated algorithms was assessed via Rand index value, with classic k-means procedure being employed as a point of reference. In conclusion it was established, KHA and FPA can be considered as being effective clustering tools.