基于XGBoost的新零售优惠券使用行为预测

为实现新零售优惠券的定向投放,提出了对用户优惠券使用行为预测的模型.该文采用XGBoost算法,突破了以TAM模型(技术接受模型)为基础解释个人优惠券使用意愿的传统方法,并基于口碑网的真实交易数据进行了特征提取和用户使用行为建模.在K折交叉验证之后通过变量重要性评分,确定了对消费者使用决策贡献度较高的特征,并与随机森林和GBDT(梯度提升决策树)算法进行了AUC(Area under curve)准确率的对比.该研究证明了基于XGBoost的集成学习算法在优惠券使用行为预测中的有效性,对新零售精准营销有重要的现实意义.

On XGBoost-Based Prediction of New Retail Coupon Usage Behavior

To achieve targeted delivery of new retail coupons,a model for predicting coupon usage behavior has been proposed.The XGBoost algorithm is adopted,which breaks through the traditional method of interpreting personal willingness to use coupons based on the TAM model.Feature extraction and user behavior modeling are formulated based on real transaction data.After K-fold cross-validation,the variable importance score was used to determine characteristics that have a significant contribution to consumer decision-making.AUC accuracy comparison with random forest and GBDT algorithm is also performed.This research proves the effectiveness of the XGBoost-based ensemble algorithm in predicting the use behavior of coupons and has important practical significance for precise new retail marketing.