基于最大流HITS的改进算法  

Improvement of HITS algorithm based on maximal flow

作　　者：刘馨月[1] 赵明砚[1] 张宪超[1] 刘芳芳[1] 

机构地区：[1]大连理工大学软件学院,辽宁大连116024

出　　处：《计算机工程与应用》2008年第17期141-143,150,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60503003; No.60673066)

摘　　要：HITS是一种经典的链接分析算法,其主要问题是容易发生主题漂移。针对这一问题,提出了一种改进的算法:MCHITS。MCHITS利用最大流算法对HITS进行改进:首先将root集扩展两层,然后将root中的结点作为种子结点通过最大流最小割算法发现以root集为中心的社区,社区中的页面作为MC-base集。实验结果表明MCHITS提高了查询结果的相关度,减少了主题漂移的发生。HITS is one of the classical link analysis algorithms,the main problem of it is the topic drift.In this paper,a new algorithm：MCHITS,which is based on maximal flow algorithm is prososed.It performs the expansion from root twice thus including pages which are link-distance two or less from at least one page in the root set,then by maximal flow-minimal cut algorithm finding a community that takes root set as the center,the pages in the community as MC-base set,The experimental results show that based on the MC-base,the iterative computation of Hits has a big improvement about the results,decreases the probability of the topic drift.

关 键 词：链接分析 HITS 主题漂移 社区 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]