基于K-prototype聚类的学生教育画像分析

Portraying Student Education based on the K-prototype Clustering Method

针对包含类别数据与数值数据的高校教育大数据,采用基于汉明距离与欧式距离混合度量的K-prototype进行聚类,并通过轮廓系数对聚类效果进行评价,与常用的k-means、DBSCAN进行对比,实验结果表明:基于K-prototype的聚类在教育大数据的处理上更具优势。基于聚类结果,采用词云图构建了4类学生教育画像。画像从基本属性、生活消费、生活规律等维度将学生群体划分为“思想积极型学生”“夜猫子游戏型学生”“规律作息型学生”“节俭好学型学生”4类,为学生教育管理人员提供相应的决策支持。

We adopt the K-prototype method based on the hybrid measurement of Hamming distance and Euclidean distance to cluster the big data for university education,including category data and numerical data,and then use silhouette coefficient to evaluate the clustering effect.Compared with popular clustering methods such as k-means and DBSCAN,we find that K-prototype clustering method has more advantages in processing educational big data.Based on the clustering results,we construct four types of student education portraits by using word cloud diagrams.According to their basic attributes,living expenses,daily routine,etc,students are categorized into four groups,including“students who think positively”,“students who are addicted to games”,“students with a regular schedule”,and“students who are frugal in life and active in study”,which may provide corresponding decision-making support for education administrators.