基于后验概率支持向量机在客户流失中的预测

Prediction support vector machine based on posterior probability in loss of customers

针对支持向量机处理不平衡样本时的缺陷,提出一种基于后验概率的支持向量机模型(PPSVM)。给出用户流失的后验概率,据此进行客户流失的预测,制定相应的政策留住客户。在UCI数据集以及移动公司客户数据上进行实验,实验结果表明,PPSVM能够避免类别模糊的样本对分类器的影响,获得比传统支持向量机更高的分类准确率,对于非确定性分类问题,稳健性较好,具有很强的实用性。

Aiming at the defect of the support vector machine（SVM）when dealing with unbalanced samples,a support vector machine model based on the posterior probability（PPSVM）was proposed.The posterior probability of the loss of users was given and used to predict the loss of customers,so as to formulate corresponding policies to retain customers.Results of experiments on UCI data set and mobile corporation customer data show PPSVM can avoid the influence of fuzzy sample categories on classifier for classification and gain the higher accuracy than the traditional support vector machine.At the same time,the PPSVM model has better robustness and strong applicability for the uncertain classification problem.