摘要
针对传统聚类算法在学习者群体划分时存在的问题,提出一种基于改进DBSCAN特征聚类的学习者类画像方法。首先,在分析学习者行为数据的基础上,采用改进的PCA-GRBM算法进行多维特征提取;然后充分利用同类学习者在答题过程中的相似性,在经典DBSCAN算法中引用以斯皮尔曼相关系数为度量的多重聚类步骤,提出改进的S-DBSCAN多重特征聚类算法;最后将其应用于学习者类画像的构建中,并对得到的学习者群体特点进行了分析。实验表明,提出的方法有效地提取了学习者类别特征,聚类结果取得了最小的戴维森堡丁指数,适合对学习者类进行画像研究。
Aiming at the problems of traditional clustering algorithms in learner group division,a learner class profile method based on improved DBSCAN feature clustering is proposed.Firsly it uses an improved PCA-GRBM algorithm for multi-dimensional feature extraction on the basis of analyzing learner behavior data.Then,it proposes an improved DBSCAN multi-feature clustering algorithm,which makes full use of the similarity of similar learners in the process of answering questions.The classic DBSCAN algo⁃rithm uses multiple clustering steps with Spearman correlation coefficient as a measure.Finally,it is applied to the construction of learner class profile,and the characteristics of learner group obtained are analyzed.Experiments show that the proposed method ef⁃fectively extracts learner's category characteristics,and the clustering result achieves the minimum DBI,which is suitable for pro⁃file research on learner class.
作者
李静
郝耀军
杨瑜
LI Jing;HAO Yaojun;YANG Yu(Department of Computer,Xinzhou Teachers University,Xinzhou 034000)
出处
《计算机与数字工程》
2022年第4期703-708,756,共7页
Computer & Digital Engineering
基金
山西省本科教学质量提升工程项目(编号:J2020291)资助。
