基于隐式反馈数据的个性化游戏推荐

Personalized Game Recommendation Based on Implicit Feedback

现有推荐系统通常采用评分、评论等显式反馈数据实现个性化推荐.然而,显式反馈数据由于在实际中难以获取或因质量问题而往往变得不可用,从而导致相关推荐算法的应用范围受到很大限制.与此相反,诸如点击行为、浏览记录等隐式反馈数据在现实中大量存在.本文提出了一种面向游戏玩家的基于隐式反馈数据的游戏推荐方法.该方法综合考虑了玩家操作次数、操作时长等隐式反馈数据及其时效性,构建了基于伪评分的玩家对游戏的偏好模型,而后通过改进了的SVD++(Singular Value Decomposition++)算法实现个性化游戏推荐.在大规模真实数据集上的实验结果表明本文提出的方法具有更高的推荐精确率和召回率.

The existing recommendation systems typically employ explicit feedbacks,such as ratings and reviews,to achieve personalized recommendations.However,since the explicit feedbacks are hard to obtain in practice or have poor quality,their applications in the recommendation field are limited.In contrast,implicit feedbacks such as click behaviors and browsing histories are widely available in reality.A game recommendation method based on implicit feedbacks is presented,to provide the personalized game recommendation for game players.This method integrates the time-based implicit feedbacks,such as the times and duration of user operations,and constructs the player's preference model based on the pseudo rating.Meanwhile,it gives the personalized game recommendation by the improved SVD++(Singular Value Decomposition++)algorithm.The extensive results of the experiment based on the real data set show that the proposed method has the higher precision and recall.