一种多样性增强的推荐列表选择算法

针对传统的推荐算法过于强调推荐的精准度导致推荐列表的同质化现象突出的问题, 提出了一种新的推荐列表选择算法DivEnhance。首先给出了推荐列表的多样性和效用值的定义; 然后将其建模为一个带约束的整数规划问题来求解, 通过一个参数的调整, 可以实现多样性和精准度的灵活控制。实验结果表明, 该算法可以在一定精准度损失的条件下, 大幅提高最终推荐列表的多样性。特别地, 在推荐一些新颖性较高的内容上, 该算法相对于传统的推荐算法具有较大的优势。

Recommendation list selection algorithm with diversity enhancement

Traditional recommendation algorithms overemphasize recommendation accuracy and homogenization phenomenon of recommendation lists is prominent. In view of this problem, this paper proposed a new recommendation selection algorithm called DivEnhance. First, it gave definition of recommender lists' diversity and utility, and then constructed a constrained integer programming model to solve the problem, through a parameter adjustment, it could realize the flexible control of diversity and accuracy. Experiment results demonstrate that the proposed algorithm can enhance the diversity of recommendation lists at the cost of a certain accuracy reduction. Specially, it outperformed other recommendation algorithms in recommending some novel items.