融入用户风险偏好的三支协同过滤推荐模型

Three-way collaborative filtering recommendation model integrating user risk preferences

基于用户的不同风险偏好特征,提出一种融入用户风险偏好的三支协同过滤推荐模型来提高推荐规则的准确性.首先,考虑用户的不同风险偏好对项目评分的影响,基于用户-项目评分矩阵定义用户关于项目的偏好概率测度,建立用户-项目偏好概率模型,从理论上证明了该模型是现有模型的推广和拓展.其次,利用决策粗糙集,推导出用户在不同风险偏好下的三支推荐模型阈值表达.然后,以上述工作为基础,将推荐准确性和推荐成本作为优化目标,设计基于粒子群优化算法的用户偏好概率模型参数确定方法 .最后,在MovieLens数据集上的实验验证了提出模型的有效性.

Based on the characteristics of different risk preferences of users,a three-way collaborative filtering recommendation model integrating user risk preferences is proposed to improve the accuracy of recommendation rules.First,considering the impact of different risk preferences of users on item ratings,a user's preference probability measure for items is defined based on the user-item rating matrix,and a user-item preference probability model is established,which is theoretically proved to be a generalization and extension of the existing model.Second,by using decision-theoretic rough set,mathematical expressions for the threshold of three-way recommendation under different risk preferences of users are deduced.Moreover,based on the aforementioned work,taken recommendation accuracy and recommendation cost as optimization objectives,a method for determining parameters of the user's preference probability model is designed based on particle swarm optimization algorithm.Finally,experiments on the MovieLens dataset verify the effectiveness of the proposed model.