Robust Planning of Non-guaranteed Targeted Display Advertising
Online display advertising are sold via event-based auctions at the spot market. Auction lacks a holistic view of the publisher’s ad resource and thus suffer from well-recognized drawbacks: on one hand, the publisher’s revenue is often not maximized; on the other hand, advertisers overall may receive less exposure. In this project, to address the above drawbacks, we propose a planning approach for ad publishers to better allocate their ad resources. To implement the approach, we propose a data-driven optimization framework with two building blocks: (1) an arbitrary-point-inflated Poisson regression model to deal with users’ ad clicking behavior, where we will directly forecast the number of clicks, instead of the click probability of an individual display as in the literature; (2) a mixed-integer nonlinear programming model to solve for the optimal ad resource allocation plan, for which we will need to develop efficient algorithms due to its large scale.
|Effective start/end date||1/10/16 -> …|