Research And Implementation Of Social Group Structure Method Based On Mobile Communication Data

With the progress of mobile technology and communication technology innovation,the telecommunications industry continues to develop and the scale of main business are growing.Due to the telecom operators'advantages,it has collected a large amount of user communication data.However,with the Internet OTT(Over The Top)business on the strong impact of mobile operators,operators urgently need to dig potential users information from The vast amounts of data,to help them provide user-centered service.Among them,the construction of social user groups based on mobile communication data can help telecom operators to accurately analyze customer social structure,and thus promote the service and improving the quality of marketing.In this paper,by using the vast amounts of user communication data,we design a set of effective social group construction algorithm.First 'of all,through preprocessing the original mobile communication records,so that the records can better reflect the user's social willingness.And then we respectively consider both of the location and contact information,so as to measure social relationships between users.A method of measuring the communication correlation is designed based on the call data of the mobile users from the time of conversation,the number of times,the frequency distribution of the call and the periodicity of the communication.For the location data of the mobile user,we design a measuring user position correlation method by establishing the user's mobile space and time sequence.In the end,the communication relation between users is abstracted as a complex network based on social relations.In this paper,a group construction algorithm based on maximal group and ant colony algorithm is designed for undirected weighted complex network.Through transforming node network as links diagram,we design a kind of group construction algorithm based on link density clustering.In order to make the algorithm designed in this paper suitable for the mass mobile communication network of the great amount of data,our algorithms are paralleled implemention based on MapReduce.As a result,the designed social group construction algorithm based on mobile phone data is compared with a series of excellent algorithms in benchmark network data and multiple real call records data sets.The experimental results show that the proposed algorithms are feasible and effective,whether on the artificial benchmark network or on the real data set.