Efficient Group Advertising over Public Display Based on User Viewing Time
Yanbin Wang, Yuchun Guo, Yishuai Chen · 2015
Public display appears everywhere providing a variety of services.Advertising is the main revenue source of public display.Existing advertising strategies, however, is inefficient, due to the difficulty of collecting users' explicit feedback, e.g., user rating or click.No implicit feedback method is proposed so far.In this paper, we propose to use users' viewing time to infer audiences' interest in the advertisement.We further design practical algorithms to handle the case that multiple users arrive and watch the public display at the same time.To derive a group's collective interest in an advertisement, we consider the media display structure characteristics of advertisements, as we found users' viewing time of public display is highly related to the advertisement's media display structure.We implemented a system prototype and conducted extensive field experiment to evaluate it.Experimental results demonstrate that users viewing time is a valid implicit feedback to infer audience preference on public display advertisement, and the proposed advertising strategies are practical and effective for group aware advertising.• It is accurate to infer a user's interest in a public display advertisement from her viewing time.• The proposed UCB-like algorithm is efficient to real-time update users' interest on advertisement and has very low computation complexity and thus fit public display efficiently.• The group advertising strategies can obtain good user advertising experience.In particular, the introduction of media similarity improves the accuracy of advertisement significantly.