Using the ELCA Technique to Find R-cliques in a Graph Database

• Main Authors: Yu-Kai Chang, 張祐愷
• Other Authors: Ya-Hui Chang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/63368965226625174999

碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 102 === Graph databases have been applied to represent complex data, and keyword search on graph databases is a convenient mechanism for users. However, the time complexity for keyword searching on a graph database is much higher than searching on tree databases. Therefore, it is an important issue to improve the efficiency on searching graph databases. [KA11] proposed the idea of ``R-clique’’ as the result of keyword search on graph databases, where an R-clique contains every query keyword and the distance among every keyword match is under a given threshold R. [KA11] has also proposed the Branch and Bound algorithm to quickly find all R-cliques in a graph database. In this thesis, we study how to convert a graph database to a tree database, and use tree search methods to generate the R-cliques in a graph database. We propose several strategies to convert a graph database into a tree database. We then use the ELCA search method to get search results efficiently. Finally, we design a series of experiments to compare the different tree converting strategies and analyze the pros and cons of each method. Experimental results show that the “Most-branched” tree has best performance in terms of efficiency and recall.