Analysis And Prediction Of Students’ Behavior Based On Campus Card Data

With the wide application of science and technology education and Internet of Things systems in colleges and universities,the construction of campus digitalization has been promoted,which provides a data basis for analyzing student behavior and mining the laws of student life.Schools have been forming various data,such as basic student information,consumption records,book borrowing records,and sports records,which reveal the daily life status of students.Over time,university databases have accumulated a large amount of data,forming a huge campus big data environment.Therefore,it is very necessary to conduct indepth mining and analysis of these data.This study uses data warehouse technology,data preprocessing technology,data mining technology,and user portraits.Based on the behavior data of students in a university in Southwest China in the 2020-2021 academic year,it analyzes students’ campus behavior from the dimensions of consumption,book borrowing,and sports.Analyze and use the poor student prediction model to predict the poverty level of poor applicants,so that student managers can more comprehensively and accurately locate different student groups,laying the foundation for the refined management of colleges and universities.The main research contents are as follows :By collecting students ’ school data,firstly,data preprocessing technology is used to perform data cleaning,integration,and conversion.Secondly,the data of students ’consumption behavior,book borrowing behavior and sports behavior in school are statistically extracted.Finally,the data warehouse technology is used to design the conceptual model,logical model,dimension table and fact table of students ’ basic information and behavior data.The data warehouse of students ’ behavior is constructed to realize the unified management and extraction of students ’ basic information and behavior data.Aiming at the division of student groups,firstly,the clustering index system of student consumption behavior,book borrowing behavior and sports behavior is designed by using the constructed student behavior data warehouse.Secondly,the K-Means clustering algorithm is improved by using the method based on density and weight.Then,the improved K-Means clustering algorithm is used to mine the behavior data of students,and the student groups under different behaviors are found.Finally,the user portrait technology is used to generate student portraits based on the mining results of various student behavior data obtained above,so as to help student management workers fully understand the living conditions of students and provide decision-making reference for university managers to formulate relevant policies.On the basis of the above,aiming at the auxiliary judgment of poor students,firstly,the statistical data of students ’ consumption behavior and the original data of poor students are extracted from the student behavior data warehouse to generate the prediction data set of poor students.Then,a Naive Bayes algorithm combining multi-class attribute weighting and orthogonal transformation is proposed,and the effectiveness of the algorithm is verified by experiments.Finally,the algorithm is used to assist the determination of poor students,and the comprehensive accuracy of the prediction of poor students is 99.93 %,which provides important decision support for the determination of poor students in colleges and universities.