Homogeneous Cluster Analysis
Mika Sato‐Ilic · Procedia Computer Science · 2018
Cluster analysis has been extensively applied in many areas, such as engineering, information science, science policy, life science, behavioral science, social science, and earth science. Currently, there is a demand for the classification of large amounts of complex data in order to both extract its latent structure represented by exploratory obtained clusters (or groups) and to summarize it. Since the central issue is this extraction and summarization, the focus of cluster analysis research has been to better capture a smaller number of clusters made up of similar data from the larger complex data then has previously been possible. However, if we observe a multiple number of datasets, and we need to obtain a single result from these datasets using cluster analysis, we must obtain clusters of similar data in a dataset while some external information in other datasets is balanced over the clusters. For example, if we consider that there are K workers and n working locations in which each location has a different working load, then we have two kinds of datasets; one is geographical data of n locations and the other is data of working load with respect to n locations. The purpose of the proposed cluster analysis is to obtain K clusters from the geographical dataset of n locations, while working loads of each cluster are homogeneous over the K clusters. In other words, we try to obtain clusters, each of which contains a balance of geographical distance and workload. Therefore, this paper proposes a cluster analysis that obtains homogeneous clusters in which each has similar data. A numerical example is given to show the good performance.