基于Bootstrap方法的电商优惠券目标人群定位分析

发布时间：2018-01-27 17:15

本文关键词： 自助法 Logit模型 ROC曲线 AUC 正例覆盖率负例覆盖率　出处：《南京大学》2015年硕士论文　论文类型：学位论文

【摘要】：互联网电商网站迅猛发展,产生了大量用户各层面的信息,大数据时代也己到来。如何利用好海量数据,怎样为电商营销服务,这类问题受到越来越多的关注。本文即是利用电商的大数据进行精准营销的案例。本文阐述的是互联网电商优惠券的发放,如何定位目标人群,使得优惠券使用度提高,从而提升优惠券的ROI。本文先是根据Logit模型利用样本进行模型拟合,从18个指标中找到相关度高的自变量指标,然后计算ROC曲线,AUC大小来估计模型的预测能力,并且利用bootstrap方法,计算AUC的方差和置信区间,并对以此加以验证Logit模型的预测能力和分类效果。最佳概率分类阀值,使得Sensitivity, Specificity最大化。并利用Bootstrap方法对Sensitivity, Specificity分别进行区间估计。
[Abstract]:The rapid development of Internet e-commerce website has produced a large number of users at all levels of information, big data era has arrived. How to make good use of massive data, how to serve e-commerce marketing. This kind of problem receives more and more attention. This paper is a case of precision marketing by using big data of e-commerce. This article describes the issue of Internet e-commerce coupons and how to locate the target population. The use of coupons is improved, thus enhancing the ROI of coupons. Firstly, according to the Logit model, we use samples to fit the model, and find the independent variable index with high correlation from 18 indicators. Then the ROC curve is calculated to estimate the prediction ability of the model, and the variance and confidence interval of AUC are calculated by using the bootstrap method. And to verify the prediction ability and classification effect of Logit model. The best probability classification threshold value makes Sensitivity. Specificity is maximized and the interval estimates of sensibility and Specificity are carried out using Bootstrap method.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F713.36;F224

【相似文献】