A scalable pattern mining approach to web graph compression with communities

摘要: A link server is a system designed to support efficient implementations of graph computations on the web graph. In this work, we present compression scheme for specifically accommodate community queries and other random access algorithms servers. We use frequent pattern mining approach extract meaningful connectivity formations. Our Virtual Node Miner achieves without sacrificing by generating virtual nodes from itemsets in vertex adjacency lists. The phase guarantees scalability bounding complexity O(E log E). facilitate global mining, relaxing requirement be sorted URL, enabling discovery both inter-domain as well intra-domain patterns. As consequence, allows incremental updates. Further, it not only facilitates but can also expedite such PageRank local walks implementing them directly compressed demonstrate effectiveness proposed several publicly available large data sets. Experimental results indicate that algorithm 10- 15-fold most real word sets