Family Relationship Recognition Based On Mobile Communication Social Network

With the improvement of the mobile communication network,the market for family customers is expanding,and it has become one of the new strategic development directions of domestic operators.However,the research on family groups in the field of mobile communication is still immature,and the related algorithm models are still relatively few,and most of them are based on mathematical statistics to describe customers'relationships,which is difficult to meet the needs of family business marketing.The traditional solution ignores the behavioral relationship between the customers,and can't effectively distinguish the similarities and differences between family members,colleagues,friends and other strong association groups.Simultaneously,the traditional solution requires a large memory resource in the processing of the base station signaling data,and the time cost is also high.Therefore,this paper has carried out in-depth research on the identification of family relationships,aiming at solving the difficulties in the traditional schemes such as low recognition accuracy,high engineering cost and limited application scenarios.Based on the mobile data realizing ability platform,this paper comprehensively analyzes the call behavior of mobile communication users and the base station signaling,fully exploits the behavioral feature patterns of family members,and characterization of the characteristics of different user groups.Then,based on the GBDT and LR,the extraction of family members candidate group is realized,which reduces the engineering cost of manually constructing a large number of feature indicators in the traditional scheme.Simultaneously,this paper designs a reasonable vector representation of the base station signaling,which uses the weighted summation of base station trajectory vectors of time-space segmentation,reducing the memory resource requirements of the server,and this paper realizes the joint training of the mobile communication customers mixed relationship network by introducing network representation learning,enhancing the accuracy of the identification of the family members.In addition,this paper analyzes the attribute characteristics of mobile communication members,and designs a family members role recognition scheme,which effectively evaluates the rationality of the family members relationship model recognition results and shortens the evaluation period of the model effect in the traditional scheme.Finally,based on the built family members relationship recognition model,this paper designs multiple business modules,and realizes the unified operation of family business marketing,which provides basic technical support for future business needs.