使用Clementine多项式Logistic回归判定电信客户类别

摘要

According to usage patterns of customers, telecom service providers classify them into four categories: basic services, e-services, additional services and full-service. Based on demographic data, we can predict group membership and recommend the appropriate service for the telecommunications potential customers. Use logistic regression train to discriminate telecom customer category. Compare the predicted classified values and the samples’ actually observed values. By calculating the statistics, we can obtain the discriminating accuracy values of different variables we choose for different category customers. Recording fields of high discriminating accuracy, we can investigate the categories of potential customers. The experimental results show that the user’s education level, job tenure, settled lives and the number of family members leads high discriminating accuracy for additional services and full-service class. These fields should be investigated from potential users’ demographics.%基于服务使用模式，电信服务提供商将他们的客户分成基本服务、电子服务、附加服务和全方位服务4类，则可以根据人口统计数据来预测组成员，为电信潜在客户推荐相应的套餐方案。使用多项式logistic回归进行类别的判定训练，通过比对样本中客户类别的实际观察值，统计获得多种协变量下各类客户的判定准确率，记录用于较高判定准确率的字段，用于考察潜在客户所对应的类别。实验结果表明：用户的受教育水平、在职年期、定居年期和家庭成员人数对附加服务类和全方位服务类的判定准确率较高，可作为判定潜在用户类别的有效特征字段。