基于RS_RBF的电信企业客户流失预测

针对电信企业客户流失的不规律性,提出以粗糙集（RS）_RBF神经网络作为电信企业客户流失的预测模型.首先利用粗糙集理论对客户属性约简,简化了网络结构.其次提出以约简后的决策表的规则支持度作为径向基函数的响应宽度基准,此种赋值方法相比传统方法更具合理性和科学性.最后利用正交最小二乘法（OLS）求得对网络输出贡献度较大的条件属性集和网络权值.把本模型与其它RBF预测模型应用于电信企业客户流失预测并且进行效果比较,实验结果证明了本模型的有效性和高效性.

Loss of Customers Forecast of Telecom Enterprise Based on RS_RBF

To the irregularity loss of customers of telecom enterprise,A rough set（RS）_RBF neural network as the loss of customers of telecom enterprise forecast model is proposed.First,It reduce the customer attributes by rough set theory,simplify the network structure.Second,proposed rule support degree after the reduction of decision table as baseline of response width of radial basis function,this assignment method is more reasonable and scientific than traditional methods.Finally,obtain the condition attribute set that with a large output contribution to the network by the orthogonal least squares（OLS）.By using this model and other RBF forecast model to the loss of customers forecast of telecom enterprise,compare the effects,This experimental show this model is more effective and efficiency.