在管理决策的制定中,分类已经成为一种十分重要的方法和技术。由于现实客户数据常常是不完整的,因此,研究不完整数据的客户分类问题具有重要意义。通过分析以往分类过程中对不完整数据的处理方法,提出了一种基于动态分类器集成选择的不完整数据分类方法DCES-ID。分别在UCI客户分类数据集以及某券商客户数据集上进行分类的实验和实证分析。结果表明,与已有的6种分类算法相比,DCES-ID算法具有更高的分类准确性及稳定性,能够更有效地进行客户分类。
Classification is an important method and technique in decision making and data mining.As the actual customer data sets are often incomplete,and therefore the study of customer classification method for incomplete data is of great significance.This paper analyses traditional methods of processing incomplete data for classification and proposes a new customer classification method based on dynamic ensemble selection for incomplete data:DCES-ID,and the new algorithm is used to brokerages customer classification.The results indicate that the DCES-ID algorithm may be more useful for customer classification because it provides higher classification accuracy and stability than the other five kinds of classification algorithm.The advantage of the new ones is proved through theoretical analysis and demonstration.