基于改进的FolkRank广告推荐及预测算法

Adapted-Folk Rank for Ads recommendation and prediction

广告的点击率预测是指利用点击日记预测的点击率,其结果受到多方面因素影响,其中包括用户性质。有效的预测广告点击率,可以提高用户对网站展示广告的满意程度。基于Page Rank的Folk Rank算法在一个用户、资源、标签的三元组中进行迭代计算,求出推荐的标签。本文使用改进后的Folk Rank方法,通过个性化地从目标节点向其他各个结点传递权重,达到广告推荐的目的中,并实现对推荐广告的广告点击率预测。

The prediction of advertising click-through rate refers to using the click logs to forecast the click-through rates, the result is affected by various factors, including the property of user. Effective prediction of ad clicks can improve the satisfaction of users about the displayed advertising. The Folk Rank based on Page Rank is an algorithm which uses a third-order tensor tuple to calculate iteratively, then there comes the recommended tags. In this paper we propose a new algorithm called Adapted-Folk Rank to personalizedly transfer the weight from preference node to the other nodes so that it can recommend Ads. Furthermore we can predict the click-through rate of recommended Ads.