The Study Of Campus Consumer Activity Pattern Of College Student Based On Smart-Card Data

With the rapid development of education informatization,big data in education is getting more and more attention.Big data techniques are used to mine hidden students behavior pattern,which gives students a comprehensive evaluation.In addition,it can not only discover the abnormal behavior of students,but also has great significance in improving the traditional student management model.The campus-card system records student's consumption activities in real time,which provides abundant database for data mining in colleges.Researchers have carried out extensive researches around them,mainly including student dietary habit analysis,abnormal behavioral discovery,financial hardship identification,friend relationship mining,etc..However,there is less quantitative research on students behavior while qualitative assessment is essential to understand the complexity of student activities.Most of quantitative approaches use psychological questionnaires designed by the self-assessment scale,but lack of effectiveness.Based on campus-card data,the quantitative evaluation of student activity patterns is quantified,objective,accurate and simple.Taking into account the student activities vary at consumption time and place,this thesis use campus-card consumption data to study the pattern of student campus activities.In terms of human dynamics,this thesis firstly analyzed the characteristics of students' consumption activities,including the distribution of time interval and time correlations.Then,in order to quantify the student's consumption activity,this thesis introduced active entropy models based on the information entropy.With setting up different entropy aggregation functions,the active entropy measures the spatial and temporal features of students' daily activities.The empirical results on real-world data demonstrate effectiveness of entropy models,and a case study has been done to evaluate the applicability about active entropy.The main work includes the following aspects:(1)From the perspective of human dynamics,the distribution and time characteristics of student's consumption sequence have been studied,and patterns about student population's activity are macroscopically grasped.The research results reveal that the distribution of breakfast consumption days for students of different grades follows the exponential distribution of different indices,and the sequence of breakfast consumption activities has a long-term positive correlation.The sequence of consumption interval time obeys the mixed distribution,exponential in the head,and power-laws for tail.Individual students breakfast displays strong burstiness and weak memory.(2)Proposed the quantification of student's campus consumption activity index-active entropy(temporal active entropy,spatial active entropy,Temporal-Spatial weighted active entropy)and corresponding algorithms.The analysis of the distribution of entropy values of different student groups was conducted.The application of true entropy predicts the number of activities of the consumer sites at different time periods.(3)Further application of active entropy combine with data mining.The test results show that active entropy and other consumption characteristics can classify students into three categories.There is a strong correlation between student active entropy,breakfast consumption,academic performance,and book borrowing.