Research On User Relationship Evaluation Model Of ’Yunnan Integrated Service Platform’

The rapid development and popularization of Internet applications in recent years made social network and online office become indispensable tools for people’s life and work,at the same time,under the policy of "Internet Plus" in our country,a new topic of "Internet Plus Party-building" and "Wisdom Party-building" are formed.To keep up closely with the Party,Yunnan Province has established a network-based integrated service platform named "Yunnan Pioneers" with online social and e-government functions.With the rapid increase of the users and the introduction of new functions,the information overload phenomenon occurs on the platform due to the explosive increase of data volume,resulting in the low accuracy and inefficient data-running on user-relationship evaluation model on the platform.Some functions based on the user-relationship evaluation model like community discovery and information dissemination will not be able to run well.Therefore,the thesis will propose an improved algorithm model based on the classical user-relationship evaluation model to ease the problem of information overload that occurs during the fast-developing period of the platform.The main work of the paper are as follows:(1)Analysis the current status of Yunnan Integrated Service PlatformAnalyzes the functions and needs of Yunnan integrated service platform,and studies the low accuracy and inefficiency of the user-relationship evaluation model encountered by the platform and other social platforms.Summed up with the common deficiencies of the current model,such as: the dimension of the evaluation index is not enough;ignoring the particularity of the platform for Party-building;using weighted summation but the influence of user preference on the result is not considered.(2)Put forward a multi-dimensional based optimization of user relationship evaluation modelStarts from the social text data,user location data,common friend data,user interaction degree data,and user background attribute data,specific improved algorithms are proposed,and similarity calculation is performed using the maximum likelihood method,proposes a multi-dimensional optimization user-relationship evaluation model(MOUEM).(3)Designed a MOUEM based community discovery method(MBCDM).Based on MOUEM,this thesis designs a method of community discovery.This method uses the idea of localized attribute table to optimize community discovery to achieve rapid community discovery for users with different attributes.Finally,the MOUEM model is applied to the community of party and group users of the Yunnan comprehensive service platform.Experiments are conducted on real user data of the platform,similarity calculation methods based on network distance,and user similarity calculation methods based on social content,etc.Compared with the algorithm,the MOUEM model has significantly improved accuracy and is more targeted.