A Clustering Method Based on the Most Similar Relation Diagram of Datasets

摘要

In this paper we present a novel clustering analysis method based on the Most Similar Relation Diagram (MSRD). MSRD is a diagram in which each datum of a dataset is linked to its most similar data. By cutting off some links in the diagram a certain number of clusters are formed. A compare of the MSRD method with hierarchical method is implemented. Clustering experiences using MSRD were done and the results demonstrated the validity of the method.