Accepted Papers

TV Advertisement Scheduling by Learning Expert Intentions

Yasuhisa Suzuki, Wemer Wee and Itaru Nishioka

This paper considers the automation of a typical complex advertisement scheduling system in broadcast television (TV) networks. Compared to traditional TV advertisement scheduling, we consider the case where not all requests for advertising slots are known at the same time, and time-consuming negotiations related to balancing the TV network’s and advertisers’ priorities have to be minimized. Although there are existing works that automatically provide schedules using mathematical optimization, the applicability of these techniques to our problem is limited due to the cumbersome formulations necessary for handling vague conditions and aesthetic domain-specific rules necessary for advertisers’ satisfaction.

To automate the system, we propose a data-driven approach that uses intention learning on top of mathematical optimization and clustering to imitate the decision-making process of scheduling experts. The scheduling of TV ads is automated via mathematical programming, and the expert objectives and constraints are learned from historical demonstrations using inverse optimization. The clustering of TV ads and the learning of associated intentions are used to deal with the cold start problem related to ad requests from new companies or products. Our proposed system is validated on actual dataset from a Japanese TV network. Experiments show that our system can more closely reproduce the experts’ schedules compared to standard optimization approaches, demonstrating the potential of our work in reducing personnel costs and improving advertisers’ experience. Based on its promising results, our proposed system is being prepared for commercial deployment.


How can we assist you?

We'll be updating the website as information becomes available. If you have a question that requires immediate attention, please feel free to contact us. Thank you!

Please enter the word you see in the image below: