Analysis Of Student Behavior Based On Parallel H-mine Algorithm

With the development of internet technology,the informational construction of colleges and universities has been continuously deepened,and the level of informational construction of smart campus,campus card service platforms and other systems has been continuously improved.In the process of campus informational construction,the college has accumulated a large amount of data related to students.Campus' management concerns on how to analyze these data and mine the behavioral characteristics to provide auxiliary decision for school and affairs' management of student.With the further deepening of campus informational construction especially,the data will continue rapid growth,how to efficiently mine and analyze the growing data is the next challenge.To solve this problem,the paper analyzes the frequent item set algorithm widely,the integration method of parallel strategy and big data mining algorithm named H-mine is studied,and experiments student's behavioral characteristics based on check-in data.First,the paper compares a variety of classic frequent item set algorithms,analyzes the usage scenarios,advantages and disadvantages,studies the process of serial H-mine algorithm deal with sparse data.Then,because the low efficiency of mine sparse big data with existing frequent item set algorithms,a load balancing parallel H-mine algorithm is proposed.The algorithm is based on the serial H-mine algorithm and the Spark platform,and has been achieved load balancing through theoretical calculation for each partition,which satisfies the behavioral analysis of student based on dynamic and batch processing check-in data.Finally,preprocesses the check-in data of student from 2016 to 2017,obtains a variety of students' behavioral data statistically,uses kmeans++ clustering algorithm and custom methods to discretize behavioral indicators,applies parallel H-mine algorithm to mine frequent item sets related to students' behaviors,then the associational analysis method is used to analyze the behavior of students with the help of frequent item sets,the correlational analysis method is used to analyze the correlation between the behaviors on the basis of associational analysis,which provides a basis for objective and comprehensive analysis of student behaviors.Experiment show that the parallel H-mine algorithm achieves load balancing on calculation for each partition,and applies the parallel H-mine algorithm to dynamic and batch processing students' data is feasible and effictive.Apply parallel big data technology to big data of campus,not only can analyze the behaviorcharacteristics of students,and then provide auxiliary decision for colleges and students' management,but also can take corresponding measures according to the behavior of students to adapt the teaching model to achieve personalized,high-quality teaching.