Clustering the Labeled and Unlabeled Datasets using New MST based Divide and Conquer Technique

Clustering is the process of partitioning the data set into subsets called clusters, so that the data in each subset share some properties in common. Clustering is an important tool to explore the hidden structures of modern large Databases. Because of the huge variety of the problems and data distributions, different classical clustering algorithms, such as hierarchical, partitional, density-based and model-based clustering approaches, have been developed and no techniques are completely satisfactory for all the cases. Sufficient empirical evidences have shown that a New Minimum Spanning Tree (NMST) representation is quite invariant to the detailed geometric changes in cluster boundaries. Therefore, the shape of a cluster has little impact on the performance of MST based clustering algorithms, which allows us to overcome many of the problems faced by the classical clustering algorithms. NMST based clustering algorithms also have the ability to detect clusters with irregular boundaries and so they are being widely used in practice. In these MST based clustering algorithms, search for nearest neighbour is to be done in the construction of NMST. This search is the main source of computation and the standard solutions take O(N) time. In our paper, we present a fast minimum spanning tree-inspired clustering algorithm. This algorithm uses an efficient implementation of the cut and the cycle property of the NMST, that can have much better performance than O(N) time. General Terms: Data Mining, Image Processing, Artificial Intelligence, Graph Theory and Design Analysis of Algorithms.