广东工业大学学报 ›› 2022, Vol. 39 ›› Issue (02): 55-61.doi: 10.12052/gdutxb.210056

• 综合研究 • 上一篇    下一篇

基于细粒度混杂平衡的营销效果评估方法

郑佳碧1, 杨振国1, 刘文印1,2   

  1. 1. 广东工业大学 计算机学院, 广东 广州 510006;
    2. 鹏城实验室 网络空间安全研究中心, 广东 深圳 518066
  • 收稿日期:2021-03-26 出版日期:2022-03-10 发布日期:2022-04-02
  • 通信作者: 刘文印(1966-)，男，教授，博士，主要研究方向为网络身份安全、大数据分析、计算机视觉、机器人、文本挖掘，E-mail：liuwy@gdut.edu.cn
  • 作者简介:郑佳碧(1982-)，女，博士，主要研究方向为因果关系、用户行为，E-mail：zhengjiabi@gmail.com
  • 基金资助:
    国家自然科学基金资助项目(62076073)；广东省基础与应用基础研究基金资助项目(2020A1515010616)；广东省创新科研团队计划项目(2014ZT05G157)

Marketing-Effect Estimation Based on Fine-grained Confounder Balancing

Zheng Jia-bi1, Yang Zhen-guo1, Liu Wen-yin1,2   

  1. 1. School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China;
    2. Cyberspace Security Research Center, Peng Cheng Laboratory, Shenzhen 518066, China
  • Received:2021-03-26 Online:2022-03-10 Published:2022-04-02

HTML

PDF (PC)

2580

摘要： 营销效果评估是精准化市场营销的重要依据。虽然因果效应估计为营销效果评估提供了有效的研究框架，但现有因果效应估计方法主要针对群体因果效应评估问题而设计，在基于细粒度个体的因果效应评估时往往面临着用户时序特征描述困难、时序非时序混合特征混杂因子选择稳定性差等问题。针对上述问题，提出了细粒度混杂平衡的营销效果评估方法。首先，引入长短记忆神经网络对用户时序特征进行建模；然后，采用稀疏多层神经网络从时序和非时序属性等混杂因子中学习样本权重；最后，采用上阶段学习到的样本权重对营销效果进行独立评估。实验结果验证了本文提出的时序特征建模和近邻匹配思想可降低效果评估的偏差、提升稳定性，对于营销策略优化具有参考意义。

关键词: 营销效果评估, 因果效应, 混杂平衡, 深度学习, 时序特征

Abstract: The marketing effect estimation is an important issue for precision marketing. Although the causal effect estimation has provided the research framework for this problem, the existing method mainly focuses on the coarse-grained causal effect estimation on a group of users. When dealing with the fine-grained evaluation problem on an individual user, the existing methods usually fail to solve the challenges raised by the user temporal feature modeling and high dimensional confounder selection. To address the above challenges, a fine-grained deep confounder balancing marketing effect estimation method is proposed. Firstly, a long and short memory network is introduced to model the users’ temporal features. Secondly, a neural network with multi-sparse-connected layers is devised to construct the confounders balancing weight network for both temporal and non-temporal features. Finally, the marketing effect is evaluated using the balancing weight learned in the previous stage.It shows that the time feature modeling and nearest neighbor matching ideas proposed in this paper can reduce the deviation of the effect estimation and improve the stability, it has reference significance for the optimization of marketing strategy.

Key words: marketing-effect estimation, causal effect, confounder balancing, deep learning, timing characteristics

中图分类号: 

  • TP391
[1] HAIR J J F, SARSTEDT M. Data, measurement, and causal inferences in machine learning: opportunities and challenges for marketing [J]. Journal of Marketing Theory and Practice, 2021, 29(1): 1-13.
[2] VARIAN H R. Causal inference in economics and marketing [J]. Proceedings of the National Academy of Sciences, 2016, 113(27): 7310-7315.
[3] PEARL J, MACKENZIE D. The book of why: the new science of cause and effect[M]. New York: Basic Books, 2018.
[4] RUBIN D B, WATERMAN R P. Estimating the causal effects of marketing interventions using propensity score methodology [J]. Statistical Science, 2006, 21(2): 206-222.
[5] 苗旺, 刘春辰, 耿直. 因果推断的统计方法[J]. 中国科学:数学, 2018, 48(12): 1753-1778.
MIAO W, LIU C C, GENG Z. Statistical approaches for causal inference [J]. Sci Sin Math, 2018, 48(12): 1753-1778.
[6] 郝志峰, 黎伊婷, 蔡瑞初, 等. 基于因果模型的社交网络用户购物行为研究[J]. 广东工业大学学报, 2020, 37(3): 1-8.
HAO Z F, LI Y T, CAI R C, et al. A research on users’ shopping behaviors in social network based on causal model [J]. Journal of Guangdong University of Technology, 2020, 37(3): 1-8.
[7] KOHAVI R, LONGBOTHAM R. Online controlled experiments and A/B testing [J]. Encyclopedia of Machine Learning and Data Mining, 2017, 7(8): 922-929.
[8] SAJONS G B. Estimating the causal effect of measured endogenous variables: a tutorial on experimentally randomized instrumental variables [J]. The Leadership Quarterly, 2020, 31(5): 101348.
[9] VARNALI K. Online behavioral advertising: an integrative review [J]. Journal of Marketing Communications, 2021, 27(1): 93-114.
[10] JOHANSSON F, SHALIT U, SONTAG D. Learning representations for counterfactual inference[C]// International Conference on Machine Learning. Cambridge MA: PMLR, 2016: 3020-3029.
[11] HARTFORD J, LEWIS G, LEYTON-BROWN K, et al. Deep IV: a flexible approach for counterfactual prediction[C]//International Conference on Machine Learning. Cambridge MA: PMLR, 2017: 1414-1423.
[12] LOUIZOS C, SHALIT U, MOOIJ J, et al. Causal effect inference with deep latent-variable models[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. New York: NeurIPS, 2017: 6449-6459.
[13] TAN Z. Regularized calibrated estimation of propensity scores with model misspecification and high-dimensional data [J]. Biometrika, 2020, 107(1): 137-158.
[14] ZHAO Q. Covariate balancing propensity score by tailored loss functions [J]. Annals of Statistics, 2019, 47(2): 965-993.
[15] IMAI K, RATKOVIC M. Covariate balancing propensity score [J]. Journal of the Royal Statistical Society, 2014, 76(1): 243-263.
[16] KALLUS N. Deepmatch: balancing deep covariate representations for causal inference using adversarial training[C]//International Conference on Machine Learning. Cambridge MA: PMLR, 2020: 5067-5077.
[17] YOON J, JORDON J, VAN DER SCHAAR M. GANITE: Estimation of individualized treatment effects using generative adversarial nets[C]//International Conference on Learning Representations. Vancouver: ICLR, 2018.
[18] KUANG K, CUI P, LI B, et al. Treatment effect estimation via differentiated confounder balancing and regression [J]. ACM Transactions on Knowledge Discovery from Data (TKDD), 2019, 14(1): 1-25.
[19] FAN Q, HSU Y C, LIELI R P, et al. Estimation of conditional average treatment effects with high-dimensional data (2019-05-06) [2020-08-15]. http://ideas.repec.org/p/arx/papers/1908.02399.html
[20] LI S, FU Y. Matching on balanced nonlinear representations for treatment effects estimation [J]. Advances in Neural Information Processing Systems, 2017, 30: 929-939.
[21] YAO L, LI S, LI Y, et al. Representation learning for treatment effect estimation from observational data [J]. Advances in Neural Information Processing Systems, Montreal:NeurIPS, 2018(14): 31-40.
[22] HAN Y, LAM J C K, LI V O K. A Bayesian LSTM model to evaluate the effects of air pollution control regulations in China[C]//2018 IEEE International Conference on Big Data (Big Data). Piscataway: IEEE, 2018: 4465-4468.
[23] SILVA D F, GIUSTI R, KEOGH E, et al. Speeding up similarity search under dynamic time warping by pruning unpromising alignments [J]. Data Mining and Knowledge Discovery, 2018, 32(4): 988-1016.
[1] 吴俊贤, 何元烈. 基于通道注意力的自监督深度估计方法[J]. 广东工业大学学报, 2023, 40(02): 22-29.
[2] 刘冬宁, 王子奇, 曾艳姣, 文福燕, 王洋. 基于复合编码特征LSTM的基因甲基化位点预测方法[J]. 广东工业大学学报, 2023, 40(01): 1-9.
[3] 徐伟锋, 蔡述庭, 熊晓明. 基于深度特征的单目视觉惯导里程计[J]. 广东工业大学学报, 2023, 40(01): 56-60,76.
[4] 刘洪伟, 林伟振, 温展明, 陈燕君, 易闽琦. 基于MABM的消费者情感倾向识别模型——以电影评论为例[J]. 广东工业大学学报, 2022, 39(06): 1-9.
[5] 章云, 王晓东. 基于受限样本的深度学习综述与思考[J]. 广东工业大学学报, 2022, 39(05): 1-8.
[6] Gary Yen, 栗波, 谢胜利. 地球流体动力学模型恢复的长短期记忆网络渐进优化方法[J]. 广东工业大学学报, 2021, 38(06): 1-8.
[7] 赖峻, 刘震宇, 刘圣海. 基于全局数据混洗的小样本数据预测方法[J]. 广东工业大学学报, 2021, 38(03): 17-21.
[8] 岑仕杰, 何元烈, 陈小聪. 结合注意力与无监督深度学习的单目深度估计[J]. 广东工业大学学报, 2020, 37(04): 35-41.
[9] 曾碧, 任万灵, 陈云华. 基于CycleGAN的非配对人脸图片光照归一化方法[J]. 广东工业大学学报, 2018, 35(05): 11-19.
[10] 陈旭, 张军, 陈文伟, 李硕豪. 卷积网络深度学习算法与实例[J]. 广东工业大学学报, 2017, 34(06): 20-26.
[11] 刘震宇, 李嘉俊, 王昆. 一种基于深度自编码器的指纹匹配定位方法[J]. 广东工业大学学报, 2017, 34(05): 15-21.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发