Recommendation diversity is increasingly becoming an important indicator to evaluate the performance of the recommendation system.There is little consideration of the dissimilarity of the item attribute value for the existing methods of improving the recommendation diversity.In this paper, an improved algorithm for recommendation diversity based on the dissimilarity of attribute value is proposed.Firstly, the dissimilarity of attribute value and item is measured.Secondly, items are clustered according to the item dissimilarity.Finally, combined with clustering information, the initial Top-N recommendation list generated by the existing recommendation algorithm is optimized.Experimental results show that the proposed algorithm can effectively improve the recommendation diversity while maintaining an acceptable level of recommendation accuracy.%推荐多样性日益成为评价推荐系统性能的重要指标.现有的提高推荐多样性方法缺少对项目属性值差异度问题的考虑.论文提出一种基于属性值差异度的推荐多样性改进算法.首先,针对属性值差异度以及项目差异度进行度量;其次,基于项目差异度进行项目聚类;最后,结合聚类信息对现有的推荐算法生成的初始Top-N推荐列表进行优化.实验结果表明论文所提出的算法在保证推荐结果准确率的同时能有效提高推荐的多样性.
展开▼