ULTR-CTR: Fast Page Grouping Using URL Truncation for Real-Time Click Through Rate Estimation
2017 IEEE International Conference on Information Reuse and Integration (IRI)(2017)
摘要
Click Trough Rate (CTR) estimation is a crucial measure (or procedure) in online digital advertising (Ad). It defines the probability of a displayed Ad being clicked by viewers, and can serve as a performance metric to validate the effectiveness of Ad campaigns with respect to pages, sites, or media types etc. Due to real-time response nature of the online digital advertising eco-systems, it is vital to accurately estimate the CTR in real-time. In this paper, we propose a URL truncation based fast page grouping for real-time CTR estimation (ULTR-CTR). Our hypothesis is that web pages under the same URL folder have similar page style and semantic content, and will share similar CTR values. While grouping web pages based on the page content is computationally expensive and hardly scalable to real-time applications, we use simple URL truncation to estimate CTR values of different site-folder combinations. Our empirical study and A/B test carried out on a commercial bidding engine confirm that ULTR-CTR based bidding achieves 2.0% performance gain in CTR estimation, and 1.4% lift in Gross Profit (GP) gain.
更多查看译文
关键词
Click Trough Rate estimation,media types,real-time response nature,online digital advertising eco-systems,real-time CTR estimation,web pages,URL folder,similar page style,similar CTR values,page content,real-time applications,simple URL truncation,ULTR-CTR based bidding,fast page grouping,site-folder combinations,commercial bidding engine
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络