Research On Correlation Analysis Of University Students' Score Based On Data Mining

With the continuous development of education in China,the number of students in college has been increasing.However,students' performance in school varies considerably.Every year,there are students failing in exams,repeating academic years or even dropping out.In order to improve students' learning quality and reduce the difficulty of teaching management,this paper uses data mining to analyze students' performance data,in order to find out the factors that affect students' performance.The results of data mining can provide personalized guidance for students and effective information for the adjustment and improvement of the teaching methods.The dataset used in this paper is the score data of students in the major of Communication Engineering in Beijing University of Post and Telecommunication.In order to complete score discretization and association rules mining,this paper conducted a thorough research on clustering algorithm and association rule mining algorithm,proposed the improved methods,and conducted a multi-angle score association analysis using the improved algorithm.The main research contents are as follows:1)The improvement of clustering algorithm for the discretization of score data.According to the distribution characteristics and discretization requirements of score data,this paper discusses the shortcomings and inapplicability of the existing discretization methods.Then,this paper proposed the Initial Cluster Center Optimized and Outliers Preprocessed K-Means Clustering Algorithm based on sample distribution density,and described the methods of calculating the sample density,initial clustering center domain and other parameters.The evaluation indexes like the Calinski-Harabasz index and Silhouette coefficient are used to measure the performance of the improved K-Means algorithm and other existing discretization methods.2)The research of association rules mining algorithm with the introduction of "Interest".For the traditional association rules mining algorithm cannot judge the interestingness of the rules,this paper analyzed several interest metrics,and adopted a differentia-based interest measure that introduces Conf(X(?)Y)as a modification.The association rule mining algorithm which introduces this measure of interest can judge the validity and value of the association rules.3)Score data analysis of curriculum dimension and student dimension.The correlation analysis of curriculum dimension validated the effectiveness of the improved clustering algorithm and the improved association rules mining algorithm by applying the algorithm to real score data,and analyzed the connection between the courses based on the mining results.The correlation analysis of student dimension used several methods like frequent pattern mining,the improved association rules mining algorithm and statistical analysis to explore the factors that affect students'performance from multiple perspectives.