[2D3-OS-7a] 広告とAI(1/2)

[2D3-OS-7a-03] Contents Allocation for Brand Recognition in TV Ads

〇Kotaro Tanahashi¹, Kengo Hamada¹, Shunji Umetani² (1. Recruit Co., Ltd., 2. Osaka University)

Keywords:mathematical optimization, machine learning, TV ads

In advertising, it is important to consider the effective frequency which is the number of times a person must be exposed to an advertisement for a brand to be recognized. In this research, we propose an integer programming model for maximizing the brand recognition in TV advertising. Specifically, the problem is formulated as an assignment problem of the several kinds of TV advertisements to given broadcasting time frames, to maximize the number of viewers whose expected number of views exceeds the effective frequency. We used the actual viewing log data as an input to our model, which was collected from the smart TVs. Since the input data is too large to solve the optimization problem in practical time, we devised a sampling method to reduce the problem size by using the solution of the LP problem. We confirmed that the model can achieve the higher score in evaluations with a proposed sampling method than the naive sampling method.