基於智能化電子看板的針對觀看人數最大化問題的排程演算法設計

Abstract

近年來，電子看板工業經歷了巨大的增長，所以基於電子看板的廣告行銷問題研究變得越來越熱。在這篇論文裡，我們對基於電子看板的廣告排程定義了一個觀看人數最大化的問題。為了幫助排程演算法的設計，我們設計並訓練了一個廣告模型，用來預測廣告的吸引力。我們提出了兩大類的廣告排程演算法：基於機率的排程演算法，和廣告對排程演算法。基於機率的排程演算法幫助我們省下很多時間去決定廣告排程的順序；而廣告對排程演算法開創性地先將廣告配對，然後把每個廣告對當作一個播放單位，再進行排程。當缺乏有吸引力廣告時，我們有創造性地主動加入一些有吸引力的內容，用以提高整體廣告的觀看人數。最後，我們使用真實數據和人造數據進行了一系列實驗。實驗結果表明我們所提的方法在觀看人數最大化的問題上非常有效。

In recent years, the digital signage industry has experienced tremendous growth, so the topic of digital signage-based advertisement scheduling become hotter. In this paper, we study a impression maximization problem for advertisement scheduling on digital signages. We train an AD-model to make our system more intelligent to help our scheduling. We proposed two sets of advertisement scheduling algorithms, probability-based scheduling and AD-Pair scheduling. Our probability-based scheduling algorithms help us to save much time to schedule the playing order. And the AD-Pair scheduling algorithms groundbreakingly bind two Ads together as a playing unit. In the situation that there are not many, even any strong Ads, we creatively add some strong Ads (we call them "Attractions") into the advertisements set given by advertisers, to attract more people to watch the signage. We have conducted comprehensive experiments on synthetic datasets, and the experimental results show that the proposed models are suitable for effective impression maximization in digital signage advertising.