Abstract: To expand new business， the telecommunication companies need to understand their users deeply. So the data of the user behavior was analyzed in the telecommunication system by using the big data analyzing technology. A clique percolation based local fitness method was proposed for weighted network algorithm (CLFMw) based on the call logs of users in the telecommunication network. The social relationships were established from all of the call connections. Based on the local fitness method (LFM) and the clique percolation method (CPM)， the user group was constructed with CLFMw algorithm. According to the massive data sets of the telecommunication system， parallelization design was used based on the MapReduce programming model. Finally， the group construction algorithm is verified by the simulation data set and the real data set for China Mobile. The experimental results show that this method is well performed but also feasible and effective.

Key words: telecommunication networks, relationship closeness, group construction, complex network, MapReduce