Imparting clusters from the tree of cores using graph based clustering

摘要

Data clustering is a data analyzing technique that groups data based on the similarity. The similarities between the objects in the same group are high when data are well clustered and the similarities between objects in different groups are low. The data clustering technique is widely used in different areas such as bioinformatics, image segmentation and market research. All of the well-known clustering algorithms require input parameters which are hard to determine but have a significant influence on the clustering result. There are many methods to find arbitrary shaped clusters. DBSCAN(Density Based Spatial Clustering with Noise) is an example of clustering algorithms. This method is able to detect communities of arbitrary size and shape. It is also used to identify hubs and outliers. However, it needs user to specify a minimum similarity ε and a minimum cluster size μ to define clusters, and is sensitive to the parameter ε which is hard to determine. To overcome this method, proposed a novel density based network clustering method called graph-based clustering (gClu). The main objective of the proposed system is to form clusters from a core connected tree with maximum accuracy and also to detect hubs and outliers.