Research On Data Mining And Student Behavior Early Warning Analysis In University Information System

With the gradual acceleration of information construction in colleges and universities,digital campuses and smart campuses have gradually become important means of scientific management in universities.They have been applied in fields such as teaching,scientific research,and student management,which have improved quality and efficiency on management.After many years of operation,various application systems in colleges and universities generate a large amount of data,which are stored in various application system databases.These data play only a partial role in their independent fields.However,the information resources of independent application systems should be integrated and form a unified resource management in digital campus and smart campus.Then,by analyzing the integrated data,the hidden price of massive data should be discovered.The potential connections between students' behavior characteristics can assist student management and teaching management in colleges and universities.Moreover,it can change traditional education models,learning lifestyles and student management models.In view of the above problems,it takes the original data of various application systems as the research data.Firstly,obtain the data from the student management system,campus card system,educational administration system and access control system.It adopts cluster analysis algorithm and correlation analysis algorithms to mining and analyze the potential connections from the massive data of the application systems.Establish a platform for student behavior analysis and prediction analysis based on Spark parallel processing through studying student behavior characteristics,analyzing student behavior patterns,clustering student behavior categories,establishing student behavior characteristics models and using experimental data to verify and enhance the model accuracy.By analyzing the relationship between the students' campus card system consumption rules and the students' academic performance of the educational administration system,it can give the suggestion and guideline for the evaluation of the Identification of poor students and outstanding students.By analyzing the access control data of student apartments and library access control systems,it can analyze the Students' behavioral trajectories in university,and then help students manage staff to analyze student behavior patterns and explore different methods of student behavioral guidance.By analyzing the network public opinion information of campus internet access log system and varioussocial networks such as the official post bar,Weibo,forum,etc.,the hotspot direction and sensitive information keywords are screened out,and the students' ideological state is grasped in time to prevent false public opinion which can be spread and even become sensational hot events that can affect social development and change.The clustering algorithm used in this research platform is based on K-Means optimization algorithm with information entropy and density improvement.The figure K is calculated in this optimal clustering algorithm.The students' behavior characteristics,such as academic achievement,card consumption record and apartment access control records are used as the clustering dimensions of K-Means.The information entropy is used to determine the attribute weight values to classify these dimensions and find out the features that are prominent in the same cluster.This research provides valuable data for student behavior analysis through the student behavior characteristics prediction and early warning platform.It can realize classification and intelligent analysis for students' behavior,provide prediction and early warning functions on students' behaviors in university,and provide significant meanings for follow-up teaching management and scientific research management.