Clustering Algorithm for Travel Distance Analysis
Nadežda Zeņina, Arkady Borisov · Information Technology and Management Science · 2013
An important problem in the application of cluster analysis is the decision regarding how many clusters should be derived from the data. The aim of the paper is to determine a number of clusters with a distinctive breaking point (elbow), calculating variance ratio criterion (VRC) by Calinski and Harabasz and J-index in order to check robustness of cluster solutions. Agglomerative hierarchical clustering was used to group a data set that is characterized by a complex structure, which makes it difficult to identify a structure of homogeneous groups. Stability of cluster solutions was performed by using different similarity measures and reordering cases in the dataset.