Algorithm Of Finding Communities Based On FCA

During the past few years, in a great number of Web community detection methods, most only focus on the plane topologic characteristics of newwork. However, seldom scholar has begun to research community's space and time dimensions. Which we care more can help user to analyze the whole structual layout and track the development process of community.The main purpose of this paper is to find communities and their hierarchy and evolution relations. Different with most traditional methods, our method wants to find community's development stages, including its ancestor and forefront, and predict its development by tracking trends, based on only the topology of network. It has no parameter, and can find turning points of communities'evolution and their network staus.Innovations of this paper are:1) Make a directed edge's end as the property of the start, and study the concept collection of start points. Based on FCA (Formal Concept Analysis) , the communities'hierarchy and interactive relations and evolution of trends can be intuitively displayed.2) Separate network into levels by the transfer of relation, and distil star topology structures from network, which are the most close and stable, and the smallest unit.3) Give a new definition of community based on sociological theory, and the smallest unit to affect the community development is the little star community rather than a single node.Improve adaptability of FCA (Formal Concept Analysis) to find communities here.4) Propose a rule extraction method based on the concept tree and a collaborative recommendation method.