COMPEL: The International Journal for Computation and Mathematics in Electrical and Electronic Engineering

Purpose – A wide number of applications requires classifying or grouping data into a set of categories or clusters. The most popular clustering techniques to achieve this objective are K-means clustering and hierarchical clustering. However, both of these methods necessitate the a priori setting of the cluster number. The purpose of this paper is to present a clustering method based on the use of a niching genetic algorithm to overcome this problem.

Design/methodology/approach – The proposed approach aims at finding the best compromise between the inter-cluster distance maximization and the intra-cluster distance minimization through the silhouette index optimization. It is capable of investigating in parallel multiple cluster configurations without requiring any assumption about the cluster number.

Findings – The effectiveness of the proposed approach is demonstrated on 2D benchmarks with non-overlapping and overlapping clusters.

Originality/value – The proposed approach is also applied to the clustering analysis of railway driving profiles in the context of hybrid supply design. Such a method can help designers to identify different system configurations in compliance with the corresponding clusters: it may guide suppliers towards “market segmentation”, not only fulfilling economic constraints but also technical design objectives.