Campus Big Data Based Analysis And Visualization Of Students' Abnormal Behaviors

In recent years,due to all kinds of student events that are thought-provoking,there is a great social concern over the increasing frequency of student problems.How to effectively identify the potential and unconventional abnormal behaviors of students on campus from the large base of wide range of activities and strong personalities has become the research focus of campus problems.With the rapid development of data storage,mining and visualization technology,a new solution has been provided for exploring abnormal.However,existing scholars mainly focus on macroscopic analysis in the mining of students' abnormal behaviors,while there are few studies on the detection and visual analysis of students' deviant behaviors.This paper is based on the students' behavior data of a college in Beijing,including the student consumption data,the student recharge data,the student library borrowing data,the student library access control data,the student network browsing data,etc.These information proposed a set of methods about the exploration,exceptions,correlation analysis and visualization of unusual behavior of students,and in this way,computers can automatically detect abnormal activities on campus.On this basis,the group correlation among abnormal individuals was measured through finding the connection inside student groups with similar behaviors on college grounds.In order to verify the reliability of the experimental results,this paper selected all undergraduate students in the whole university as a sample for experiments,and obtained prior abnormal student group information from the designated guidance personnel.This paper mainly includes the following three aspects of research content:(1)Individual detection of student anomalies based on unsupervised clustering.In this paper,the characteristics of life behavior of campus students are extracted and expressed.The unsupervised classification of student groups based on life behavior is realized by clustering method,and the relative abnormal factors are defined to detect the possible abnormal individuals.This method was used to conduct abnormal behavior retrieval of the data of undergraduates in 2017 academic year(N=8,685).Together with the student behavioral information received the experiment conducted qualitative analysis on individual behaviors,and verified them by asking relevant counselors,and finding out that students expressed abnormal behaviors such as long-term truancy,leaving school and irregular campus activities.The experimental results show that this method can provide accurate judgment for campus managers to retrieve individuals with abnormal behaviors.(2)Correlation analysis of student groups based on the optimal degree of module.Firstly,three kinds of operators for measuring student similarity are proposed to calculate the behavioral similarity of any two students in the undergraduate range of the university.The accuracy of the proposed operators for measuring spatial and temporal similarity is as high as 87%.The following step uses the spatial and temporal similarity index,in which the index is used as a student's feature,as between individuals,together with using the optimal algorithm of group modeling to analyze student association degree,finally,in a situation when the modular can get to the maximum degree by mining data,we get the student group which has the biggest similarity.The validation of a relevant counselor,who is going to confirm that the gathered data is accurate is the final step.The experiment group showed that correlation is more common between students who share the same bedroom,couples and close friends.The method can be used to search abnormal population according to abnormal individual detection results.(3)Visualization of abnormal behaviors of students.The abnormal behaviors on campus are analyzed through the visualization methods such as feature description,space-time correlation and abnormal individual exploration.The spatial and temporal similarity operators are used to construct the graph of chord connection and uniform connection among students and measure the spatial and temporal correlation among them.By using relative anomaly factors and space-time correlation between individuals,abnormal individual graphs and correlated abnormal graphs were constructed to explore abnormal individuals and groups existing on the campus.Combined with the results of the first two parts,this visualization method verified the abnormal individual students with few social relations and isolated life obtained in the first part,and the groups with close correlation obtained in the second part.This method provides a new method for the correlation analysis of abnormal individuals and groups.This paper constructs an exploration method for the abnormal behaviors of individuals and student groups.By using the campus life data,this paper explores the abnormal behaviors on the campus by means of individual detection of student abnormalities,analysis of association inside student groups and visualizing it.The research results can provide more scientific theoretical support and more accurate decision support for the monitoring and management of abnormal behaviors on campuses.