针对目前客户流失预测方法的不足,在利用训练样本中不同类个数比值来确定各类惩罚参数的基础上,改进标准的C-支持向量分类机(SVC)。通过以美国某电信公司客户流失预测为实例,与标准C-SVC、人工神经网络、决策树、贝叶斯分类器等方法进行了对比,发现该方法能获得较好的正确率、命中率、覆盖率和提升系数.是研究客户流失预测问题的有效方法。
Aiming at the shortcomings of the methods for customer churn prediction, developed the improved C-support vector classifier( SVC ) by using the ratio of different classes in training set to evaluate the penalty parameters of the classes. It pointed out that the method could acquire the better accurate rate, hit rate, covering rate and lift coefficient, compared with normal C-SVC, aritifical neural network, decision tree, logistic regression, naive Bayesian classifier etc by predicting customer churn for some US telecommunication carrier. The results indicate that the method can be an effective measurement for customer churn prediction.