Research On Method And Application Of College Student's Behavior Analysis Based On Data Mining

With the development of new technologies such as big data and artificial intelligence,the method of applying big data mining technology in the education is attracting increasing attention,which can improve the level of management of schools.At present,the campus construction of major universities has been from the days of digital campus construction into the era of smart campus construction.Campus application systems,such as the campus card system and educational administration system,should be integrated.Then,by analyzing the collected historical data,the hidden price of massive data should be discovered and used to improve school management.Based on this background,this paper starts from the campus card of consumption records,borrowing records of book and student's score from various application systems as the foundation.Then,we conduct application analysis and research on them by using data mining and big data technology.Related research work covers the following aspects:A cluster analysis on student's consumption behavior.Firstly,the clustering characteristics of the constructed student behavior,such as expense,times,intensity and frequency of each month,were statistically analyzed to preliminarily understand the level of students' consumption ability and rules.Then the distributed K-means++ algorithm is used to divide students into three categories and analyze the clustering results.The experimental results show that using clustering characteristics such as the consumption intensity and frequency to assist school finance department which carry out accurate poverty alleviation work is feasible.At the same time,it provides a solution to find “needy students” and “fake needy student”.A correlation analysis and association rule mining analysis on student's behavior.Pearson correlation coefficient and distributed FP-Growth algorithm based on the Spark is used to analyze feature correlation on consumption records,borrowing records of book and student's score.The experimental results show that the student's number of breakfast,lunch and dinner has a strong affinity.Simultaneously,other rules have been discovered.Understanding these rules enables the school to take a more reasonable and scientific way to guide students and to introduce more humane,diversified,scientific and reasonable implementation guidelines.Research on student's social relations.The co-occurrence network of students was constructed through the campus card consumption records.Then,the LINE algorithm is used to extract student's vectors of social relationship from the co-occurrence network and study the relationship between the strength of social relationships and other behavioral characteristics of students.It is found that the closer the relationship between students,the smaller the difference in performance.Simultaneously,other rules have been discovered.At the same time,visual technology is used to display the student's social network,which can quickly find the lonely.It can help school administrators.The prediction model of the repast number was constructed.Because the repast number has time series characteristics,it can build a repast number of prediction model based on LSTM algorithm.Different lengths of time periods were used to predict the repast number with comparative analysis of experiments.The prediction model based on LSTM algorithm can achieve good results with a minimum average absolute error of 10.2 and an average accuracy rate of 80%,which can better help school cafeteria managers to dynamically make related adjustments in advance.Establish a platform for efficient student behavior analysis based on Spark parallel processing through campus data.Then the mining results are displayed and analyzed with front-end and back-end technologies based on Echarts,Bootstrap and Django.It can implement student's personal portraits,student's group portraits and early warning function modules,which provides application services of students and school administrators with a dynamic and comprehensive understanding of student's life and learning.The main contributions and innovations of this paper are as follows:Design student evaluation index from three aspects of life,learning and social.Then,establish feature library for student through student-generated data,analyzing student behavior patterns,providing a basis for objective and comprehensive evaluation of students.At the same time,the analysis results will timely feedback to the school's corresponding management department to optimize management,which can improve decision-making ability of school and student's service quality.Aim at the problems that most universities have not established a long-term dynamic funding system for needy students and have few means to identify lonely students.Machine learning algorithms such as clustering and LINE are used to analyze student's consumption data and build student's portraits and relationship networks.At the same time,a student behavior mining analysis and service platform based on Spark be set up,which can automatically detect abnormal situations of individual students and generate predictions.