Research And Application Of Clique Search In Community Detection Based On User Trajectory

The Internet industry has entered the era of mobile.With the development of intelligent mobile phone and network technology,all sorts of network applications emerging in an endless stream,such as mobile video,instant messaging,online social networks.People urgently need to connect to a high-speed network anytime and anywhere for enjoyment.Therefore,the concept of Wireless City is proposed.It provides the wireless network services in the coverage of the whole city and it has become a trend of city construction at present,known as the "5th public infrastructure".This paper combines the thought of social networking and trajectory data analysis.Puts forward the community detection problem in Wireless City based on access log.Hope to be able to mining implicit community structure by co-occurrence behavior between users based on wireless city access log,so as to create new value.After analysing the characteristics of access log,this paper adopts a series of preprocessing method on the data to reduce the calculation of user spatio-temporal similarity,so that the traditional frequent-item-sets mining algorithm can be applied to this scenario.But frequent-item-sets mining algorithm has defects in the application,therefore,this paper presents a community detecting method based on undirected weighted graph called Clique Search.The main idea of Clique Search is to map the log data to graph space to find the group that every internal nodes has connection to each other,thereby performing community detecting.The performance analysis implies that,compared with the traditional algorithm,this algorithm has significant enhancement in time efficiency and space efficiency.At last,we use some real raw data in the algorithm comparison experiment,the experiment shows that compared to the traditional algorithm,Clique Search algorithm can reduce the time consumption and space consumption,it can also complete the calculation in a constant-level time.At the same time,different threshold parameters in this algorithm represent different similarity levels in human activity,make the results applicable to different scenarios.