Clustering And Anomaly Detection Of Campus Behaviors Of Students Based On Psychological Attention

Campus behavior analysis of students is of great significance to process management in universities.Campus behaviors of university students are usually personalized and diverse.Traditional methods are difficult to characterize such behaviors uniquely and discover process information.Depending on network access logs,smart card records and Wi Fi logs,psychological attention is introduced to explore the clustering and anomaly detection of campus behaviors in the thesis.The main research contents are as follows:(1)Extraction of the attention features of multivariate campus behaviors of university students.Guided by psychological attention,the behavioral process of students is regarded as the production,distribution and flow of their individual attention on some objects.So,a Psychological Attention(PA)for campus behaviors is proposed,which can be used to evaluate the stability,breadth,distribution and transferring of attention campus behaviors.Clustering experiments show that psychological attention features have advantages over conventional statistical features in SC,CHI and DBI.The results show that psychological attention features can better distinguish students’ campus behaviors,and can support behavior clustering and anomaly detection effectively.(2)A multi-domain features based time-series clustering algorithm(STF-FCM)is proposed and verified on 5 UCR datasets.Experiments show that,compared with the traditional FCM algorithm,STF-FCM has advantages in 4 clustering evaluation indices.STF-FCM is also applied to the clustering of campus behaviors of students.It is found that there are certain differences in the evolution rules of attention patterns of different students,though behavior patterns of the most students are relatively stable in the whole semester,and a small number of students has great volatility.(3)A TCN and LSTM based approach named TL-BAD is proposed for the detection of abnormal behaviors of students.Experiments show that,at the end of the semester,the attention patterns of the behaviors of the abnormal groups really change greatly.Experiments also found that the academic performances of students are correlated with their attention models.Finally develop the visualization system.