A Mail Community Partition Algorithm with Correction

This paper discussed the mail community partition and the property of the mail community, and proposed an algorithm, called ACCD/C, to partition mail communities based on adjusting community center dynamically with correction. The algorithm adopted a method of cosine measure to evaluate the similarity between two mailboxes based on the characteristics of mailboxes´ communication behavior, partition communities by the way of adjusting the community centers dynamically, and adjust the partition result based on the characteristics of the community members. Experiments carried on real dataset show that the behavior of sending and receiving mails between mailboxes within a large LAN shows itself remarkable characteristics of clustering. Simultaneously, the results also show that the algorithm is reasonable.