基于原型的EM朴素贝叶斯模型在直接营销中的应用

摘要

In light of the problems in direct marketing the retail chain enterprises have, this paper present the naive Bayesian classification model based on EM clustering according to model mining and selection method of prototype theory. Through the experiment, it proves that this model clearly outperforms the K-means clustering naive Bayesian model and no-clustering naive Bayesian model in performance of customers purchasing behaviour prediction. At last, this model is also used in checking up the effectiveness of classification prediction conducted on new customers in direct marketing.%针对连锁型零售企业直接营销中的问题,基于原型理论的挖掘模型选择方法,提出基于EM聚类朴素贝叶斯模型,通过实验证明了该模型在客户购买行为的预测性能上明显优于基于K-means聚类朴素贝叶斯模型和无聚类的朴素贝叶斯模型.最后,利用该模型检验了直接营销中的对新客户进行分类预测的有效性.