摘要
以电信行业为应用对象,建立了一种基于多准则神经网络(MCNN)与分类回归树(CART)的的异动客户识别系统.该系统首先用多准则神经网络对客户属性进行约简,然后构造用于识别异动客户的分类回归树.通过对浙江省电信系统的大客户数据的实际验证,结果表明该系统具有较好的鲁棒性和有效性.
As the telecom system an application object, this paper presents a kind of unusual customers recognition system based on multi\|criteria neural networks (MCNN) and classification and regression tree (CART) in telecommunication system. The system reduces customers' attribute by multi\|criteria neural networks first. Then it constructs classification and regression trees to recognize unusual customers. Through actual validating to large customer data in Zhejiang telecom system, the results indicate that this system is more robustious and more effective.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2004年第5期78-83,共6页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(79970037)
