Research And Application Of Abnormality Early-Warning Of Student Compus Activities Based On Big Data Mining

With the advancement of "Modernization of Chinese Education 2035" and "Action Plan of Teaching Informatization 2.0","artificial intelligence","big data","Internet of Things" and other technologies,is widely used in building digital smart campus to help schools improve the quality of teaching and the campus’ management efficiency.In general,students’ activity safety and academic performance are the core of campus management and home-school joint education.Analyzing historical records or students’ activity data of a single network application can not achieve high accuracy results for detecting abnormal behavior and adapt to the change of student groups.Based on the question of students’ campus activities abnormality prediction which is feedback by school leaders and head teachers from secondary schools,this thesis finds "difficult to collect full quantitative activity data" and "early warning algorithm can not adapt" are the two key problems by deep research in two schools.And then the "companion" data acquisition method is proposed,which can effectively collect students’ activity data of various network applications and offline activities on campus to create a big data warehouse.Third,we propose to use "multiple linear regression","multinomial logistic regression" and “dynamic adaptive model training” to predict abnormal behavior of students.At last,the solution is tested in two secondary schools.Results indicate that our model can efficiently predict activity abnormalities of students base on "companion" data acquisition method,and can self-adaptive to the change of student groups year by year.Which is a typical application of engineering technology in engineering management to solve front-line management problems.