基于机器学习的展示广告点击率预测研究
Research on Click-Through Rate Prediction in Display Advertising Based on Machine Learning
DOI: 10.12677/HJDM.2019.92008, PDF,    科研立项经费支持
作者: 张芝悦, 黄 浩：对外经济贸易大学信息学院，北京
关键词: 展示广告；计算广告；点击率预估；机器学习； Display Advertising； Computing Advertising； Click-Through Rate Estimation； Machine Learning
摘要: 展示广告是网络广告的重要组成部分。在展示广告投放前对其点击情况进行预测不仅能够减少广告投放的成本也能够提高互联网公司的资源利用效率从而增加收入。随着大数据以及机器学习技术不断成熟，越来越多的公司采用相关技术预测广告点击率。本文从特征重要性以及模型适合性两个方面研究展示广告点击率预测问题。首先，文章通过对比广告特征、用户及上下文特征、媒体特征三大类特征发现广告特征对于广告点击率预测问题最为重要，同时加入媒体特征以及用户上下文特征也能够提升模型效果。其次，本文对比研究了常用于广告点击率预估的机器模型优劣，主要从模型性能以及模型耗时两个维度进行比较。本文发现逻辑回归模型、随机森林模型、梯度提升决策树模型是最适合解决广告点击率预测问题的机器学习模型。
Abstract: Display ads are an important part of online advertising. Predicting clicks before a display ad can not only reduce the cost of ad serving but also increase the efficiency of Internet companies’ resource utilization and increase revenue. As big data and machine learning technologies continue to mature, more and more companies are using technology to predict ad click-through rates. This paper studies the display ad click rate prediction problem from two aspects: feature importance and model suitability. Firstly, the article finds that the advertising characteristics are most important for the advertisement click rate prediction problem by comparing the characteristics of advertisement, user and context, and media characteristics. At the same time, adding media features and user context features can also improve the model effect. Secondly, this paper compares the advantages and disadvantages of the machine model commonly used in advertising click rate estimation, mainly from the two dimensions of model performance and model time consumption. This paper finds that the logistic regression model, the random forest model, and the gradient lifting decision tree model are the most suitable machine learning models for solving the problem of advertising click rate prediction.
文章引用：张芝悦, 黄浩. 基于机器学习的展示广告点击率预测研究[J]. 数据挖掘, 2019, 9(2): 60-67. https://doi.org/10.12677/HJDM.2019.92008

参考文献

[1] 艾瑞咨询. 2018年中国网络广告市场年度监测报告–简版.
http://report.iresearch.cn/report/201808/3264.shtml
[2] Richardson, M., Dominowska, E. and Ragno, R. (2007) Predicting Clicks: Estimating the Click-Through Rate for New Ads. International Conference on Word Wide Web, ACM, 521-530. [Google Scholar] [CrossRef] 
[3] 肖垚, 毕军芳, 韩易, 董启文. 在线广告中点击率预测研究[J]. 华东师范大学学报(自然科学版), 2017(5): 80-86+100.
[4] Dave, K. and Varma, V. (2010) Predicting the Click-Through Rate for Rare/New Ads. Center for Search and Information Extraction Lab International Institute of Information Technology, Hyderabad.
[5] He, X., Pan, J., Jin, O., et al. (2014) Practical Lessons from Predicting Clicks on Ads at Facebook. Proceedings of the Eighth International Workshop on Data Mining for Online Advertising, ACM, 1-9.
[6] Bagherjeiran, A., Hatch, A., Ratnaparkhi, A., et al. (2010) Large-Scale Customized Models for Advertisers. IEEE International Conference on. Data Mining Workshops (ICDMW), 1029-1036. [Google Scholar] [CrossRef] 
[7] Zhang, W., Du, T. and Wang, J. (2016) Deep Learning over Mul-ti-Field Categorical Data. European Conference on Information Retrieval, Springer, Cham, 45-57.
[8] Cheng, H.T., Koc, L., Harmsen, J., et al. (2016) Wide & Deep Learning for Recommender Systems. Proceedings of the 1st Workshop on Deep Learning for Recommender Systems, ACM, 7-10. [Google Scholar] [CrossRef] 
[9] Qu, Y., Cai, H., Ren, K., et al. (2016) Product-Based Neural Networks for User Response Prediction. IEEE 16th International Conference on Data Mining (ICDM), 1149-1154. [Google Scholar] [CrossRef] 
[10] Xiao, J., Ye, H., He, X., et al. (2017) Attentional Factorization Machines: Learning the Weight of Feature Interactions via Attention Networks. arXiv:1708.04617. [Google Scholar] [CrossRef] 
[11] Zhou, G., Song, C., Zhu, X., et al. (2017) Deep Interest Network for Click-Through Rate Prediction. arXiv:1706.06978.
[12] 施梦圜, 顾津吉. 基于平衡采样的轻量级广告点击率预估方法[J]. 计算机应用研究, 2014, 31(1): 33-36+39.
[13] 李航. 统计学习方法[M]. 北京: 清华大学出版社, 2012: 55-60.

为你推荐