摘要
提出了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。在粒度空间中采用相似度阀值进行调整粒度的粗细问题。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。
A cluster algorithm of spatial grain based on association rules of Matrix weighting is proposed.This algorithm utilizes eigenvector matrices of the document to extract its similarity degree, and then clusters to search for the frequent items of similar relation on the basis of above-mentioned association rules.Thereafter threshold value of similarity degree will be employed to adjust the area of granularity in spatial grain.Subsequently the results of the Matrix Weighting algorithm will be clustered.Experiments suggest that the precision of this algorithm excels the Apriori and the K-mean while processing documents of middle and small size. As for large-scale documents, the Time Complexity of this algorithm is lower than that of K-mean.
作者
李泽军
LI Ze-jun （hunan Institute of Technology, Hengyang 421002, China）
出处
《电脑知识与技术》
2010年第01Z期259-261,共3页
Computer Knowledge and Technology
基金
湖南教育厅科学研究基金项目（08C248）
湖南教育厅科学研究基金项目（09C297）
关键词
关联规则
粒度
聚类算法
频繁项集
association rules
granularity
cluster algorithm
frequent items
