基于梯度提升树模型的网络优惠券使用预测

互联网与实体经济融合发展背景下,网络优惠券往往承担了提升用户体验、促进再次消费的重要功能。构建梯度提升树、随机森林等模型,预测网络优惠券使用行为;并对影响因素的重要性进行排序。结果表明:梯度提升树算法的五折交叉验证平均测试精度、曲线下面积值分别为0. 804与0. 886,高于随机森林与单棵决策树算法。优惠券折扣率对于用户使用优惠券行为起着决定性影响,用户经常活动的地点离该商户最近门店的距离、领取优惠券时间等特征对用户使用优惠券行为具有重要影响。

Prediction of Internet Coupon Use Based on Gradient Boosting Decision Tree Model

The online business and the offline business of physical stores are being more closely integrated. Internet online coupons can play a role in improving the user experience and promoting re-consumption. Gradient boosting decison tree and random forest model are built to predict the use of internet coupons and rank the importance of influencing factors. The results show that the average test accuracy and area under curve value of gradient boosting decision tree algorithm are 0. 804 and 0. 886 respectively, which are higher than those of random forest and decision tree algorithm. the discount rate of the coupon plays the most important role in use of coupons. The distance between the place where the user often moves and the nearest store of the business, the day on which the coupon is received have an important influence on the use of coupons.