平滑资料整理:
实现中的一些可能问题:
-
冷启动: 初期的数据不全,导致无法制定合理计划;这可以通过数据积累,或者调整预测的节奏和步伐,进行一定程度的缓解。
-
效果评测:作为广告平台来说,可以参考整体的CPC的变化趋势,是否比之前更加平稳;对于DSP来说,可以将流量做AB测试,保持和一定比例的流量完全不打开Pacing。
-
预算无法耗尽:算法有时候对于一些Campaign的Pacing,可能导致预算无法耗尽,一个简单粗暴的方法,就是设置一天只有20个小时消耗时间,后面4个小时可以在一定程度上放水。
http://wnzhang.net/share/rtb-papers/linkedin-pacing.pdf
https://zhuanlan.zhihu.com/p/20799487
http://wnzhang.net/share/rtb-papers/throatling-pacing.pdf
- From 0.5 Million to 2.5 Million: Efficiently Scaling up Real-Time Bidding by Jianqian Shen et al. ICDM 2015.
- Smart Pacing for Effective Online Ad Campaign Optimization by Jian Xu et al. KDD 2015.
- An analytical solution to the budget-pacing problem in programmatic advertising by Joaquin Fernandez-Tapia. Working paper.
- Adaptive Targeting for Online Advertisement by Andrey Pepelyshev, Yuri Staroselskiy, Anatoly Zhigljavsky. Machine Learning, Optimization, and Big Data 2015.
- Real Time Bid Optimization with Smooth Budget Delivery in Online Advertising by Kuang-Chih Lee, Ali Jalali, Ali Dasdan. ADKDD 2013.
- Budget Pacing for Targeted Online Advertisements at LinkedIn by Deepak Agarwal et al. KDD 2014.
- Frequency Capping in Online Advertising by Niv Buchbinder et al. WADS 2011.