MuGRAM: An approach for multi-labelled graph matching

Graph mining has been a widely studied domain over the years. Graph representation of real world problems has enabled the development of simple solutions bringing in better clarity. Graph mining has various sub domains among which graph matching is a prominent one having a number of algorithms. With the rise of new applications involving large sets of networked data, the performance of these algorithms has become important. The graph based representations for social networks and communication networks have led to the evolution of multi-labelled large graphs which are still not completely handled by the existing algorithms. The requirement of a fast and efficient indexing process so as to accommodate dynamic graphs without having to opt for incremental indexing is another major challenge. In this paper, we propose MuGRAM - a multi-labelled graph matching approach aimed at addressing the above mentioned issues. This approach is capable of handling multiple labels for vertices as well as edges of reference graphs. An enhanced indexing method proposed in the paper ensures a fast indexing process. A breadth first search oriented spanning tree along with a novel technique for neighbourhood matching ensures fast query processing. Experimental evaluation of MuGRAM in comparison with some of the recent algorithms in the field highlights its superior performance.