In　this　paper,　we　introduce　a　new　incremental　algorithm　to　solve　the　unsupervised　clustering　problem　for　network　services,　based　on　attractive　force.　Inspired　by　the　intermolecular　forces,　in　this　algorithm,　we　regard　each　data　object　as　an　adaptive　agent.　Under　the　attractive　force　between　every　data　objects,　clusters　begin　to　diffuse　and　merge　until　finally　reaching　a　steady　state.　The　experiment　on　UCI　data　sets　and　data　sets　of　network　service　proves　the　validity　of　the　algorithm.