基于矩阵加权关联规则的核模糊聚类

提出一种基于矩阵加权关联规则的空间粒度聚类算法。该算法核心思想是根据文档特征向量矩阵提取文档的相似度,再在该关联规则算法上进行聚类来寻找相似关系的频繁项集。通过引入核函数,样本点被非线性变换映射到高维特征空间进行聚类,提高聚类性能。通过矩阵加权关联规则算法进行聚类。通过实验表明,在处理中小型文档时,该算法的精确度优于传统Apriori算法和K-mean算法;在处理大型文档时,该算法的时间复杂度小于传统的K-mean算法。     

基于图的关联规则改进算法

关联规则挖掘是数据挖掘研究的最重要课题之一。基于图的关联规则挖掘DLG算法通过一次扫描数据库构建关联图,然后遍历该关联图产生频繁项集,有效地提高了关联规则挖掘的性能。在分析该算法基本原理基础上,提出了一种改进的算法—DLG#。改进算法在关联图构造同时构造项集关联矩阵,在候选项集生成时结合关联图和Apriori性质对冗余项集进行剪枝,减少了候选项集数,简化了候选项集的验证。比较实验结果表明,在不同数据集和不同支持度阈值下,改进算法都能更快速的发现频繁项集,当频繁项集平均长度较大时性能提高明显。     

基于双阈值Apriori算法和非频繁项集的关联规则挖掘方法

针对从本文数据集中的正负关联规则挖掘问题,提出一种基于双阈值Apriori算法和非频繁项集的挖掘方法。首先,对通过逆文档频率(IDF)对语料库中的项(项集)进行加权,筛选出前N%的项集。然后,通过提出的双支持度阈值Apriori算法来提取频繁项集和非频繁项集,以此降低非频繁项集的数量。最后,通过置信度和升降度阈值的判断,分别从频繁项集和非频繁项集中挖掘正负关联规则。其中,创新性的利用了非频繁项集来挖掘正负关联规则。在一个医学文本数据集上的实验结果表明,提出的方法能够有效挖掘出正负关联规则,且能够大大降低项集和规则数量。     

区间值属性数据集关联规则挖掘算法仿真

对区间值属性数据集进行挖掘,可以有效分析出数据之间的关系。针对现有数据挖掘方法未对大规模数据进行聚类,导致挖掘过程占据内存大,挖掘精度低的问题,提出了一种新的区间值属性数据集挖掘算法。对问题定义、数据准备、数据提取、模式预测和数据聚类等模块进行详细分析,完成区间值属性数据聚类。根据聚类结果,将区间值属性数据分成多个数据集,挑选出能够支持最小支持度的项目集,将这些项目集作为频繁项集,进而提取出数据集之间的关联规则,将关联规则融入数据计算步骤,完成数据挖掘。为验证算法效果,进行仿真,结果表明,相较于传统挖掘算法,所提挖掘算法占用容量更小,挖掘精度更高。     

基于频繁项集树的时态关联规则挖掘算法

针对目前时态关联规则研究中存在的挖掘效率不高、规则可解释性低、未考虑项集时间关联关系等问题,在原有相关研究的基础上,提出一种新的基于频繁项集树的时态关联规则挖掘算法.通过对时间序列数据进行降维离散化处理,采用向量运算生成频繁项集,提高频繁项集挖掘效率.考虑到项集之间的时态关系以及树结构的优势,提出一种新的频繁项集树结构挖掘时态关联规则,其挖掘频繁项集与树结构构建同时进行,无需产生候选项集,提高了规则挖掘效率.实验表明,对比于其他算法,所提出算法在挖掘效率和规则解释性方面效果更好,具有较好的应用前景.     

基于激励的关联规则的挖掘

基于支持度的关联规则挖掘算法无法找到那些非频繁但效用很高的项集,基于效用的关联规则会漏掉那些效用不高但发生比较频繁、支持度和效用值的积（激励）很大的项集。提出了基于激励的关联规则挖掘问题及一种自下而上的挖掘算法HM-miner。激励综合了支持度与效用的优点,能同时度量项集的统计重要性和语义重要性。HM-miner利用激励的上界特性进行减枝,能有效挖掘高激励项集。     

基于特征加权的分布式大数据相关性挖掘方法

针对传统方法存在计算时间较长,任务分配均匀程度较差的问题,提出基于特征加权的分布式大数据相关性挖掘方法.对软子空间进行聚类,根据特征加权的不确定性表示加权聚类中心,并求解权值.设计特征选择的技术框架对特征加权进行选择,依据特征空间搜索机制完成特征筛选.根据特征筛选结果运用MapReduce编程模型对数据簇的聚类中心进行反复扫描,计算样本到聚类中心的距离,去除其中的孤立点.利用Shuffle均衡分组机制计算频繁项集,开始新项的FP树建立及频繁项集挖掘,直至完成所有频繁项集的挖掘.实验结果表明,所提方法的挖掘时间低于传统方法,并且任务分配均衡性较高,说明上述方法具有一定的应用价值.     

基于粒计算Web文档聚类

提出了一种基于粒计算Web文档聚类（WDCGrc）方法。该方法通过TF-IDF法则计算文档词条的权值,采取设定文档阈值和平均权值相结合的方法实行降维,抽取出每篇文档的主干词;建立了文档的主干词和二进制粒之间的转换,提出了基于粒计算提取文档间的关联规则算法来获取文档间的频繁项集,由频繁项集形成初始聚类,使用优化算法对初始聚类进行优化,得到最终聚类结果。实验结果表明,该方法切实有效,聚类质量较好。     

基于逆向搜索的关联规则更新算法

对现有关联规则更新算法中的增量式更新算法进行分析,发现在决策者优先关注最大频繁项目集的情况下,该算法不能以较少的数据库遍历次数快速获取最大频繁项集。针对该算法的不足,提出一种基于逆向搜索的方式进行关联规则更新的算法。该算法生成新增项集的所有频繁项集,通过将其中最大频繁项集跟原项集中最大频繁项集进行拼接、修剪,从中获得更新后的最大频繁项集。实例结果表明,该算法既降低了关联规则更新过程中对数据库的遍历次数,又实现了优先获取最大频繁项目集。     

图像纹理特征挖掘*

借助数据挖掘方法在图像中的应用,提出了一种利用图像降阶结合基元模式匹配对纹理特征进行挖掘的新思路。采用关联规则挖掘算法对图像纹理的频繁模式进行挖掘,通过联合关联规则来表达纹理。实验结果显示,挖掘出的关联规则不仅能够表达规则纹理,而且能够较好地表达随机纹理。     

一种交互式关联规则和频繁项集的可视化方法

关联规则中频繁项集数量庞大的问题是关联规则可视化要解决的一个主要问题,本文介绍了一种基于平行坐标系和项目分类树的频繁项集和关联规则可视化方法。首先,在频繁项集中设置显示边界,利用频繁项集的闭包特性,实现对大的频繁项集的剪枝;然后,结合overview+detail的视点控制技术,通过交互,由用户选择感兴趣的某一节点上的频繁项集,在de-tail窗口中详细显示,从而实现人机交互的频繁项集和关联规则可视化。     

负关联规则挖掘与特征词抽取融合的局部反馈查询扩展

针对现有信息检索系统中存在的词不匹配问题,本文提出一种基于负关联规则挖掘与特征词抽取融合的局部反馈查询扩展算法。该算法首先从前列n篇初检局部文档中抽取特征词,建立特征词库;然后,对特征词库挖掘同时含有查询词和非查询词的频繁项集和非频繁项集,由此挖掘前件是查询项的负关联规则,提取负关联规则的后件作为负关联特征词,计算负关联特征词与原查询的相关性,根据相关性在特征词库中删除负关联特征词,将余下的特征词作为最终扩展词,和原查询组合成新查询实现查询扩展。实验结果表明,该算法能有效地提高和改善信息检索性能。     

一种新的动态频繁项集挖掘方法

频繁项集挖掘是关联规则挖掘的重要步骤。在数据动态变化的环境下进行关联规则挖掘具有重要的现实意义。提出一种动态频繁项集挖掘算法,该算法建立在前一阶段挖掘的基础上,能避免过多地扫描数据库而影响挖掘性能,在最后生成全局频繁项集时,不需要全程扫描数据库,根据之前挖掘结果有选择地扫描相关的事务子集。实验表明,该算法挖掘性能远远优于Apriori算法,能有效地实现在数据动态变化环境下的挖掘频繁项集。     

Decentralized Association Rule Mining on Web Using Rough Set Theory

A central part of many algorithms for mining association rules in large data sets is a procedure that is to find so called frequent itemsets. The frequent itemsets are very large due to transactions data increasing. This paper proposes a new approach to find frequent itemsets employing rough set theory that can extract association rules for each homogenou.s cluster of transaction data records and relationships between different clusters. This paper conducts an algorithm to reduce a large number of itemsets to find valid association rules.     

Pincer-search: an efficient algorithm for discovering the maximumfrequent set

Discovering frequent itemsets is a key problem in important data mining applications, such as the discovery of association rules, strong rules, episodes, and minimal keys. Typical algorithms for solving this problem operate in a bottom-up, breadth-first search direction. The computation starts from frequent 1-itemsets (the minimum length frequent itemsets) and continues until all maximal (length) frequent itemsets are found. During the execution, every frequent itemset is explicitly considered. Such algorithms perform well when all maximal frequent itemsets are short. However, performance drastically deteriorates when some of the maximal frequent itemsets are long. We present a new algorithm which combines both the bottom-up and the top-down searches. The primary search direction is still bottom-up, but a restricted search is also conducted in the top-down direction. This search is used only for maintaining and updating a new data structure, the maximum frequent candidate set. It is used to prune early candidates that would be normally encountered in the bottom-up search. A very important characteristic of the algorithm is that it does not require explicit examination of every frequent itemset. We evaluate the performance of the algorithm using well-known synthetic benchmark databases, real-life census, and stock market databases     

负关联规则增量更新算法

讨论负关联规则的更新问题。与正关联规则增量更新不同,负关联规则不仅存在于频繁项集中,更多存在于非频繁项集中。针对该问题提出一种负关联规则增量更新算法NIUA,利用改进的Apriori算法以及集合的性质挖掘出频繁、非频繁项集和负关联规则。实验结果表明,该算法是可取的。     

Generating a Condensed Representation for Association Rules

Association rule extraction from operational datasets often produces several tens of thousands, and even millions, of association rules. Moreover, many of these rules are redundant and thus useless. Using a semantic based on the closure of the Galois connection, we define a condensed representation for association rules. This representation is characterized by frequent closed itemsets and their generators. It contains the non-redundant association rules having minimal antecedent and maximal consequent, called min-max association rules. We think that these rules are the most relevant since they are the most general non-redundant association rules. Furthermore, this representation is a basis, i.e., a generating set for all association rules, their supports and their confidences, and all of them can be retrieved needless accessing the data. We introduce algorithms for extracting this basis and for reconstructing all association rules. Results of experiments carried out on real datasets show the usefulness of this approach. In order to generate this basis when an algorithm for extracting frequent itemsets—such as Apriori for instance—is used, we also present an algorithm for deriving frequent closed itemsets and their generators from frequent itemsets without using the dataset.     

Pruning and visualizing generalized association rules in parallel coordinates

One fundamental problem for visualizing frequent itemsets and association rules is how to present a long border of frequent itemsets in an itemset lattice. Another problem comes from the lack of an effective visual metaphor to represent many-to-many relationships. This work proposes an approach for visualizing frequent itemsets and many-to-many association rules by a novel use of parallel coordinates. An association rule is visualized by connecting items in the rule, one item on each parallel coordinate, with continuous polynomial curves. In the presence of item taxonomy, each coordinate can be used to visualize an item taxonomy tree which can be expanded or shrunk by user interaction. This user interaction introduces a border, which separates displayable itemsets from nondisplayable ones, in the generalized itemset lattice. Only those itemsets that are both frequent and displayable are considered to be displayed. This approach of visualizing frequent itemsets and association rules has the following features: 1) It is capable of visualizing many-to-many rules and itemsets with many items. 2) It is capable of visualizing a large number of itemsets or rules by displaying only those ones whose items are selected by the user. 3) The closure properties of frequent itemsets and association rules are inherently supported such that the implied ones are not displayed. Usefulness of this approach is demonstrated through examples.     

On-line generation association rules over data streams

In order to efficiently trace the changes of association rules over an online data stream, this paper proposes a method of generating association rules directly over the changing set of currently frequent itemsets. While all of the currently frequent itemsets in an online data stream are monitored by the estDec method, all the association rules of every frequent itemset in the prefix tree of the estDec method are generated by the proposed method in this paper. For this purpose, a traversal stack is introduced to efficiently enumerate all association rules in the prefix tree. This online implementation can avoid the drawbacks of the conventional two-step approach. In addition, the prefix tree itself can be utilized as an index structure for finding the current support of the antecedent of an association rule. Finally, the performance of the proposed method is analyzed by a series of experiments to identify its various characteristics.