社团挖掘的粒计算方法研究

Study on community mining based on granular computing approach

挖掘社会网络中的社团结构是当今一个非常有挑战性和发展潜力的研究领域。粒计算理论具有模拟了人类思考和解决问题方式的一种智能信息处理模型,它通过将问题进行粒化成具有不同粒度大小的粒度空间上,以简化问题的求解。借助粒计算思想,改进了对结点之间的紧密关系的度量方法。基于粗糙集模型的粒计算方法给出了覆盖粒和覆盖粒度空间等概念,提出一种基于粒计算方法的社团挖掘数据模型,得到几个相关性质,并设计相应的基于粒计算社团结构挖掘方法,即将社团挖掘问题转换成在覆盖粒度空间上的粒度转换问题。通过实例在3个标准测试数据集上测试,并对比试验结果,表明所提出的方法是有效的。

Mining the community structure of social network is a challenging research area with great de- velopment potential nowadays. Granular computing（ GrC ） is an emerging computing paradigm of intelligent information processing, which simulates the ways of human thinking and solving problems. It can simplify the solution of a problem by granulating the prob｜em and implementing it in a granulation space with different levels. First of all, with the idea of granular computing, the measurement of the close relationship between nodes is improved. Next,with some concepts such as coverage granules and a corresponding cov- erage granular space being defined, a data model for mining community structure based on granular com- puting approach is built and several related properties are obtained. On the basis of the coverage granular spaces ,the method for mining community structure is presented, which converts solving the problem of community mining into realizing corresponding granularity translation in different levels in the coverage granular space. Finally,a real instance is illustrated and the tests and comparisons in three real standard data sets are implemented. Experimental results show that the proposed method is feasible.