基于SQL的图相似性查询方法

图作为一种表示复杂信息的数据结构，被广泛应用于社交网络，知识图谱，语义网，生物信息学和化学信息学等领域.随着各领域应用的普及和深入开展，如何管理这些复杂图数据是目前图数据库技术面临的巨大挑战.图的相似性查询是图数据管理中的热点问题之一.对图查询问题的研究主要包括图的相似性查询等.本文重点研究基于编辑距离（Graph Edit Distance）的图相似性查询处理问题.首先，通过对目前代表性的问题求解算法分析发现，其提出的过滤规则都具有自己的优缺点和适用性.其次，针对已有方法在过滤阶段自身存在优缺点和适用性的问题，提出一种全新的面向关系型数据库的过滤框架，新的过滤框架可以支持所有已有的过滤规则，从而通过结合不同的过滤规则来优化图相似查询算法以提高查询效率.该方法可以最大程度保留不同过滤规则的优点并克服其缺点，从而对不同查询具有普遍适用性.最后，基于PubChem数据集，通过比较算法在求解查询结果的时间消耗，验证本文提出算法的高效性及可扩展性，实验结果表明，本文提出的方法优于现有算法.

SQL-Based Solution for Fast Graph Similarity Search

Graphs are widely used to model complicated data in many areas, such as social networking, knowledge base, semantic web, bioinformatics, and cheminformatics, etc. More and more graph data are collected while it bas become a rather big chanllenging problem to manage such complex data. The database community has had a long-standing interest in querying graphdatabases, and graph similarity search is one of most popular topics. In this paper, we focus on the graph similarity search problem with edit distance contraints. Firstly, we investigate serveral state-of-the-art methods, and find that all the proposed pruning rules have limitations. None of them could outperform others on various queries. Considering this problem, we then propose a novel approach intended to support the graph similarity search inthe framework of query evaluation, using the relational model. The proposed approach developed a novel unified filtering framework by combing all the existing pruning rules. It could avoid limitations on existing pruning rules, and have more widely applications. We also conduct a series of expriments to evaluate the proposed approach. The results show that our approach could outperform all existing state-of-the-art methods.