基于评分矩阵局部低秩假设的成列协同排名算法

协同过滤方法是当今大多数推荐系统的核心.传统的协同过滤方法专注于评分预测的准确性,然而实际推荐系统的推荐结果往往是项目的排序.针对这一问题,将排名学习领域的知识引入推荐算法,设计了一种基于评分矩阵局部低秩假设的成列协同排名算法.选择直接使用计算复杂度较低的成列损失函数来优化矩阵分解模型,并通过实验验证了其在运算速度上的显著提升.在3个实际推荐系统数据集上,与当下主流推荐算法的比较实验结果表明,该算法具有良好的性能.

Listwise Collaborative Ranking Based on the Assumption of Locally Low-Rank Rating Matrix

Collaborative filtering (CF) is the core of most of today's recommender systems. Conventional CF models focus on the accuracy of predicted ratings, while the actual output of recommender systems is a list of ranked items. In response to this problem, this research introduces technologies in the field of learning to rank into recommendation algorithms and proposes a listed collaborative ranking algorithm based on the assumption that the rating matrix is locally low-rank. It directly uses list-wise ranking loss function to optimize the matrix factorization model. Significant improvement on operation speed is achieved and verified by experiment. Experiments on three real-world recommender system datasets show that the proposed algorithm is a viable approach compared with existing recommendation algorithms.