摘要
提出一种基于Bayesian信念网络(BN)的客户行为预测方法.通过知识学习构建客户行为Bayesian网络(CBN),根据CBN对预实例计算联合分布概率,准确预测了一对一营销优化中的客户行为.CBN学习算法包括连线和定向部分,复杂度为O(N4)条件相关测试.在零售行业一对一营销实际应用表明,CBN学习算法较现有BN学习算法更快构建CBN,预测精度高于朴素Bayesina分类法.
A new customer behavior prediction approach based on Bayesian belief network is presented. The customer behavior Bayesian network （CBN） is constructed through knowledge study, and joint probabilities are calculated with this network to precdict the customer behavior. The CBN learning algorithm is composed of connecting and directing parts, and complexity is O（N^4） conditional dependence test. The empirical applications in retail one-to-one marketing show that CBN is constructed more quickly by using this approach than other existing BN learning algorithm, and the accuracy is better than that of naive Bayesian classificiation.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第6期626-631,共6页
Control and Decision
基金
国家杰出青年科学基金项目(60425310)
国家863计划项目(2006AA04Z172)
