Social Tie's Mining And Evolution Law Of College Students Based On Campus Smart Card Data

The social ties of college students generally originate from campus communication behavior.Its formation is the process from quantitative to qualitative change of this behavior.Good communication behavior is the basis of establishing good social ties,which in turn guide communication behavior and promote the formation of a good learning and living atmosphere.However,how to obtain social ties of college students and evaluate them quantitatively has always been one of the research topics in the education field.At present,the mining of large-scale real-time college students' social ties is one of the hot issues.Promoted by the "Big Data Industry Development Plan(2016-2020)",the construction of Smart Campus has received widespread attention.The massive campus behavior data generated by the smart campus system is a comprehensive reflection of college students' learning and living conditions.The study based on this data is an effective way to help us understand the law of students' behavior and makes it feasible to obtain large-scale real-time social ties.The Campus Smart Card System is an important part of smart campus system.Utilizing the massive data generated by students' campus behavior to mine hidden campus social networks is of great significance to study the social law of college students.The data of Campus Smart Card System has unique characteristics:(1)in the spatial dimension,the scope of student activities is small and concentrated;in the time dimension,the distribution is extremely unbalanced.(2)Under the influence of class,dormitory and other inherent factors,the homophily of students' communication behavior is obvious.This dissertation proposes a method of mining college students' social ties based on multiple hypothesis-testing,and further proposes an intimacy hierarchical model and a hierarchical encounter model to improve the mining method and enhance the effectiveness of social ties' mining of college students.Based on these mined campus social networks,the evolution process of these social networks is analyzed.The main contribution of this dissertation can be summarized as follows:1)We propose a method of mining college students' social ties based on multiple hypothesis-testing.Visiting the same location at the same time(short time interval)is defined as a co-occurrence.The co-occurrence is a potential factor of forming social tie.Based on the statistics of co-occurrence,a method of social ties mining based on multiple hypothesis-testing(MSVC)is proposed.Firstly,we prove that the number of co-occurrences obeys Poisson distribution.Secondly,the edges between students are established based on co-occurrence.Thirdly,a test statistic is constructed.Then,the null hypothesis that co-occurrence is due to random factor is proposed,and the hypothesis-testing is carried out for each edge.If the validation result obeys this hypothesis,the co-occurrence is caused by random factors.Otherwise the co-occurrence is driven by social tie.Finally,a campus social network is constructed.2)We present an intimacy hierarchical model to eliminate the influence of spatial-temporal agglomeration on the mining of social ties among college students.Limited by the concentration of free time and fewer campus activity venues,college students' campus behaviors mainly concentrate in the three time periods of morning,noon,evening and a few popular campus activity venues.Therefore,their activity trajectories have obvious spatial-temporal agglomeration.In order to eliminate this influence,a social tie mining method of eliminating the influence of spatial-temporal agglomeration is proposed.Firstly,an adaptive timeslot method based on Gauss mixture model is used to cluster check-in points to solve the problem of selecting optimal timeslot.Secondly,the contribution factor is obtained according to the contribution of spatial-temporal factors to the formation of social ties.Then,the location weight and grouping weight based on information entropy are introduced.Finally,an intimacy hierarchical model(ESIHM)is proposed,which is used to mine college students' social ties and intimacy.Compared with the results of the questionnaire,the mined social ties are in good agreement with the real ones.3)We also propose a hierarchical encounter model based on analogous association analysis to eliminate the influence of homophily on the mining of social ties among college students.Influenced by some inherent external factors,the communication time and opportunity between students in the same group(same major or grade,etc.)are much higher than those of students in different groups.The homophily of college students' communication behaviors leads to the overestimation of social ties mining within the same group,while the cross-group social ties are difficult to be mined.To solve this problem,a hierarchical encounter model based on analogous association analysis is proposed to eliminate the influence of homophily.First,the sliding timeslot method is used to obtain the number of co-occurrences.Second,the adaptive threshold method is used to solve the problem that the threshold is difficult to set in traditional association analysis.Third,a hierarchical encounter model is constructed,and the same-and cross-group social ties are mined separately.The results show that this method eliminates the influence of homophily and improves the mining performance of social ties among cross-group students.4)We analyze the evolution of campus social network.On the basis of eliminating the influence of spatial-temporal agglomeration and homophily,we construct four campus social networks with school-year as a time window.Using complex network metrics,the evolution process of social ties are analyzed at the global and local levels.Firstly,the formation of small world phenomenon shows that any student can establish indirect social ties with almost all other students in a few steps.Secondly,by analyzing the distribution of the number of students' social ties on different attributes,it is found that girls' desire to make friends changes from stronger to weaker than that of boys.Then,the giant community gradually differentiates into relatively stable small communities in senior grades,which reveals the polarization of social ties:from fragile to stable or from fragile to extinct.Thirdly,the campus social networks always have strong college homophily and gender homophily,but the evolution trends of intra-group social ties of different genders are quite different Finally,the evolution of 3-motifs shows that the formation of two-way social ties is a critical transitional phase in the dynamic evolution from unclosed triangular ties to closed triangular ties.