排序方式: 共有271条查询结果,搜索用时 19 毫秒
11.
该文基于传统的PageRank链接分析原理,分析了PageRank在页面主题内容分析方面的不足之处,结合传统的基于内容的VSM文本分析模型.提出了一种基于向量空间模型的主题算法,并通过实验对改算法的性能进行分析。 相似文献
12.
13.
从Web结构挖掘的角度出发,比较了基于链接结构分析的PageRank和HITS 2个经典算法,针对HITS单纯利用链接,忽略主题相关性问题,利用模糊关系的合成,得到页面与查询词之间的模糊隶属关系,对原有的HITS算法进行改进.实例验证了算法的有效性. 相似文献
14.
陈学进 《计算机技术与发展》2009,19(5)
网络结构挖掘是以超链接分析为基础,从链接结构中获取有用的知识,利用这些知识,重新组织结构,使内容逻辑结构更加合理.深入研究现有的网络结构挖掘系统,并在对其核心算法PageRank和HITS中所存在的问题作了详细分析的基础上提出了自己的改进算法,主要是对每个网页定义这三个参数:PageRank,Authority,Hub,并进行分析与优化,以便得到更好的查询结果,最后设计了一个改进网络结构挖掘系统原型,根据实验结果进行分析. 相似文献
15.
Nutch是一个Java实现的开源搜索引擎。针对目前Nutch对中文进行单字切分且没有实现PageRank计算的缺点,改进PageRank算法,设计并实现基于MapReduce的PageRank计算方法,对Nutch中文分词进行改进,加入JE中文分词器。实验结果表明,改进后的Nutch具有更高的查询结果准确率和中文网页排序效果。 相似文献
16.
现有选择性计算机性能评价方法主要使用基准程序评价方法,基准程序中各子程序的输出往往因为单位不同而无法进行进一步数据处理,同时基准程序评价方法广泛采用的权重和评分方法缺乏理论依据。针对该问题,提出基于佩奇排名(PageRank)的计算机性能评价方法,采用比较数据序列间相似性的方法产生邻接矩阵从而为各项评估功能计算PageRank得分。实验结果表明,该方法能客观反映目标计算机系统的性能。 相似文献
17.
18.
19.
基于主题特征和时间因子的改进PageRank算法 总被引:2,自引:0,他引:2
经典PageRank算法单纯地考虑到对网页的链接结构进行分析,而不能考虑到网页在搜索主题方面的相关性和权威性,以及用户对新旧网页的依赖程度的不同.针对经典PageRank算法存在的上述缺陷,综合网页的主题特征和时间特征两个因素,提出了一种改进的PageRank算法WTPR(weighmd topic PageRank).该算法通过网页链接分析和内容分析来解决网页的权威程度和相关程度,通过时间因子实现PageRank值随时间的变动而浮动.仿真结果表明,改进后的算法与PageRank算法相比获得了更好的效果. 相似文献
20.
In distributed object computing (DOC) containers, cache strategy as a passive approach improves the performance by caching the recently accessed objects. By the advent of large-scale enterprise applications of DOC, caching methods fail to keep pace with the increasing importance of performance and the increasing scale of DOC system. Prefetching is an effective approach to improve the performance. It generates and stores pages or objects in caches in advance, by predicting the requests. In the current DOC container, prefetching strategy is not supported and seldom studied in the literature. The immune system with faster secondary response and its affinity network inspire applying the immune mechanisms to build holistic model of DOC for performance improvement. In this study, a co-evolutionary affinity network (CEA-Net) is proposed for prefetching distributed objects. In CEA-Net, objects are antibodies and computation tasks are antigens. Invocation relations among object classes are modeled by the immune network of antibody genotypes. Multiple affinity measures among antibody, antigen, genotypes of antibody and antigen, genotype set and antibody population are defined to model the complex relations among distributed objects. Especially in the antibody population, immune principles including clonal proliferation, immune memory, immune toleration and elimination are designed to add evolutionary features to the antibody population. Based on CEA-Net, the prefetching architecture for DOC is built including 5 main procedures, Network Abstractor, Access Recorder, Object Factory, Cache Engine and Prefetch Engine. Finally, the experimental study shows the promising access performance and the evolutionary features of CEA-Net. CEA-Net is instructive for the future design of high performance DOC containers, such as WebSphere Application Server and BEA WebLogic Application Server. 相似文献