摘要：Aiming at improving current clustering algorithms for their failure to effectively represent high-dimensional data, this paper provides a novel clustering algorithm-GMC-based on graph matching with data objects being represented as the attributed relational graph and the graph matching degree being the standard of similarity measurement. In the algorithm, graphs for classification will be matched with character pattern atlas, and classified into the class with the biggest similarity. The accuracy and rationality of this algorithm is always kept with continuous renewal of character pattern atlas. In addition, compared with the classical K-means clustering algorithm and Newman fast algorithm, this algorithm shows its own superiority and feasibility in applications of data mining.
