摘要： 在教辅系统使用过程中，儿童角色存在类别数目未知、交互状态调研数据量大、噪音多的问题。为此，提出一种基于K-均值系统聚类融合算法的交互角色挖掘方法。通过数据清理和数据变化预处理初始数据，利用K-均值聚类对角色实现初始聚类，使用动态系统聚类的方法对初始聚类结果进行处理，得到角色的分类结果。实验结果表明，该方法的分类情况较好、效率较高。

关键词: 预处理, 聚类融合, 角色, 约束, 数据清理, 数据挖掘

Abstract: In order to solve this problem that the number of character classes is unknown, and the large number of data in process of children using computer-education system, this paper presents a kind of method based on K-means Hierarchical Clustering(KMHC) algorithm to dig the children’s roles of the interaction. This method preprocess the data by using data cleaning with constraints and data changes with rules. After preprocess, this method uses K-means clustering method of interactive children in supplementary roles in the process of initial clustering, and uses the hierarchical clustering method to classify the initial results. Experimental results show that the classification of this method is better, and has good efficiency.

Key words: pretreatment, clustering fusion, role, constraint, data cleaning, data mining

中图分类号: 

• TP18