基于免疫算法的移动通信用户信用度评估研究  

Credit evaluation for mobile telephone customers based on immune algorithm

作　　者：杨宗长[1] 徐继生[2] 孙洪[2] 

机构地区：[1]湖南科技大学信息与电气工程学院,湘潭411201 [2]武汉大学电信学院,武汉430072

出　　处：《电子测量与仪器学报》2009年第8期105-110,共6页Journal of Electronic Measurement and Instrumentation

基　　金：广东移动公司重点科技(编号:GMCC19901-02)资助项目;湖南科大科学研究基金(自然科学)(编号:E50811)资助项目

摘　　要：针对移动通信中的欠费问题,依据移动用户行为属性,提出了基于人工免疫算法的用户信用度预测评估方法。该方法被应用于广东省移动通信一分公司两月全网近40万用户的信用度预测评估中,取得了全网预测评估精度为82.0%的良好结果。实验结果表明对于全网该两月的信用预测评估,提出的基于人工免疫算法的信用度预测评估方法整体上稍优于BP神经网络。同时基于决策树的期望受益分析表明,以信用度评估为基础的客户管理是一种可推广的措施或策略,并可为其他应用领域提供参考借鉴。Most mobile telephone corporations in China are puzzled by the arrears problem. Based on artificial immune algorithms, a credit measure using customer＇s behavior attributes is defined and proposed to mobile telephone customers. The proposed measure gives good results and performances a little better than a BP neural network, when it is applied to the credit evaluation for 400000 customers in a Communication Branch of Guangdong in August and Sep- tember. And the correct evaluation rates for all customers is about 82.0%. The experimental results and decision tree analysis of expected benefits show that the management for the arrears problem on the basis of client credit evaluation is a considerable measure for applying, and others could benefit from ours.

关 键 词：移动通信 用户信用度 人工免疫算法 决策树 期望受益分析 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]