基于改进动态GCN的大学生日常校园行为管理及预测研究  

Management and Prediction of Daily Campus Behavior of College Students Based on Improved Dynamic GCN

作　　者：王茜 Wang Qian(Wuhu Institute of Technology)

机构地区：[1]芜湖职业技术学院

出　　处：《哈尔滨师范大学自然科学学报》2024年第6期52-59,共8页Natural Science Journal of Harbin Normal University

基　　金：2021年度安徽省高校人文社会科学研究重点项目“人工智能时代大学生思想政治教育精准化研究”(SK2021A0926);2022年芜湖职业技术学院校级教研项目“种子花开辅导员党建思政培育工作室”(2022fdygzs01)

摘　　要：在大学校园中,学生的日常行为涉及课堂学习、社交活动、校园安全等方面,其管理与预测对于提升学校管理效率和改善学生体验至关重要.首先从校园卡、校园WIFI、课程系统、在线平台和监控系统等多个渠道收集学生行为数据,然后融合多源数据,采用动态语义特征提取和时间卷积层捕捉行为的时序规律,提出了一种新型多片段动态语义预测模型.实验结果表明,该模型在实际应用中表现出较高的准确度和鲁棒性.其分类准确度最低可达73%,最高可达85%.学生正常行为的预测值可达97%,异常行为的预测发生率最低可达3%.精确率最高为90.5%,召回率最高为91.3%,F1值最高为90.9%.由此可知,研究所提模型在大学生日常行为预测管理中表现较优,能够为该领域的技术发展提供一定的理论支持.On university campuses,students'daily behaviors involve all aspects of classroom learning,social activities,campus safety,etc.,and their management and prediction are crucial to enhance the efficiency of school management and improve students'experience.In this paper,students'behavior data from multiple channels such as campus card,campus WIFI,course system.online platform and monitoring system is first collected.Then integrated multi-source data,used dynamic semantic feature extraction and time convolution layer to capture the time sequence rule of behavior,a new multi-fragment dynamic semantic prediction model is proposed.The experimental results show that the model has high accuracy and robustness in practical application.The experimental results show that the classification accuracy of the model can be as low as 73%and as high as 85%.The predicted value of students'normal behavior can be up to 97%,and the predicted incidence of abnormal behavior can be as low as 3%.The maximum precision is 90.5%,the maximum recall is 91.3%and the maximum F1 value is 90.9%.It can be seen that the proposed model of the study performs better in the prediction and management of college students'daily behaviors.And some theoretical support for the development of technology in this field are provided.

关 键 词：日常行为 预测 大学生 动态语义 多片段 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]