BP神经网络技术在移动通信客户信用分析中的应用

BP神经网络技术在移动通信客户信用分析中的应用 An Application of BP Neural Network Technology on the Credit Analysis for Mobile Clients 陈俊艳 CHEN Jun-yan 1 2 first-author 徐继生 Xu Ji-sheng 1 2 李湛 Li Zhan 1 2 武汉大学电子信息学院 武汉大学电子信息学院 School of Electronic Information, Wuhan University School of Electronic Information, Wuhan University 武汉大学电子信息学院,湖北,武汉,430072 武汉大学电子信息学院,湖北,武汉,430072 提出了一种移动通信客户信用预测评估的方法.该方法基于移动通信客户行为属性的统计分析,以其作为客户信用预测评估的依据,建立 BP神经网络的训练和识别的模型.通过随机选择 2003 年 8 月全网数据（403 206个客户）中3%作为样本进行训练,确定了网络模型参数.与已有的信用评估方法相比,该方法的泛化性能高,适用于大规模的客户信用评估,所得到的客户特征属性对移动通信客户信用研究提供了理论和实践的依据. A new technology of credit forecast and evaluation for mobile clients is presented based on the statistic analysis of client behaviour character in this paper and a system is built to train and recognize client credit state by means of BP neural networks.With many experiments and training on data which are 3% of data in Aug, 2003,network model paraments are gained. Being compared with other methods on credit evaluation, the model have better generalization ability, can be specially used for massive credit evaluation.The characters being gained have better reference value for research of credit for mobile clients. 移动通信 客户行为 客户信用度 模式识别 BP神经网络 mobile communication client behavior client credit pattern recognition BP neural networks F626 国家自然科学基金资助项目（49984001）;广东移动通信有限责任公司重点科技项目 2005-01-01 2021-04-01 1