A challenging issue in emerging Peer-to-Peer (P2P) systems is the efficient locating appropriate resource within the great amount of available resources. To group peers with similar cooperative services into the same community appears very promising. However, most of the existing P2P systems group peers statically under the assumption of steady topology during runtime. This paper focuses on a method to form community topology by self-organization of peers based on interests for P2P systems. As peers interests change, peers dynamically adjust their configuration and move to the right community gradually. The experiment proves that the employment of selforganization mechanism leads to a flexible as well as scalable community formation technique.