一种面向社会化标签的并行双聚类算法

A Parallel Biclustering Algorithm for Social Tagging

随着社会化网络的快速发展,社会化数据呈现爆炸式增长,挖掘社会化数据的局部信息成为有效利用社会化数据的研究热点。相对于传统聚类方法,双聚类能够更好地挖掘社会化数据中的局部信息。较高的计算复杂度成为使用双聚类挖掘大数据集中局部信息的关键问题。通过对几何双聚类产生过程的研究与分析,提出了一种改进的并行几何双聚类方法。该方法通过过滤子双聚类合并过程中产生无效的子双聚类,降低算法的计算量,而且利用多核计算机的优势,使用并行算法,从而提高双聚类算法的效率。

With the rapid development of social networks,a great number of social data can be acquired. Extracting the local information has become the focus in the research of social data. Compared with traditional clustering method,biclustering can better exploit the local information of the social data. However,the computational complexity of biclustering is high,which is the bottleneck of mining the local information. Based on researching and analyzing the process of generating biclustering,propose an improved parallel geometric bicluster-ing method. By filtering the invalid biclustering which was generated in the process of combining sub-bicluster,the complexity can be re-duced. Moreover,by using the advantages of multi-core processors and the parallel algorithm,can improve the efficiency of biclustering on social data.