Abstract: The characteristics of the real social network were analyzed， and a network model with multi-relationships network between nodes was established. Based on this， the concept of influence of interaction between nodes was defined， and a community discovery algorithm for multi-relationship network (CDMN) was proposed. The rationality of the presented definitions was verified theoretically， and a series of experiments compared with other algorithms were conducted. Experimental results showed that the proposed community discovery algorithm for multi-relationship networks has higher accuracy and lower time complexity than other classical algorithms， and has important research significance and practical value.

Key words: online social network, multi-relationship network, community discovery, heterogeneous network, overlapping community