用于评价推荐系统的多样性指数的研究

Research on Diversity Index for Evaluating Recommendation System

针对当今数据量的庞大导致用户获取所需信息困难以及推荐系统评价体系缺乏多样性评价指标的问题,提出基于三部图校准的Herfindahl多样性指数,通过该指标来量化推荐系统的多样性。首先,根据设定好的分类方式进行URL分类;进而设计形成“类别URL用户”的三部图;其次,对原本的Herfindahl指数进行改良,减少数量的差异对多样性的影响;最后,结合改良的Herfindahl多样性指数,得到推荐系统的多样性指数。多样性指数的出现有助于在评价推荐系统时,不仅关注推荐的准确与否,而且考虑推荐信息是否全面。实验表明,基于此实验提出的方法所得的改良后的Herfindahl指数可以对推荐系统类别受众多样性进行准确的量化。

Aiming at the problem that the huge amount of data makes it difficult for users to obtain the required information and the recommendation system evaluation system lacks diversity evaluation indicators,the Herfindahl diversity index based on the tripartite graph calibration is proposed to quantify the diversity of the recommendation system.First,URLs are classified according to the set classification method;then a three-part graph of category-URL-user can be designed and formed;Secondly,the original Herfindahl index is improved to reduce the impact of quantitative differences on diversity;Finally,Combined with the improved Herfindahl diversity index,the diversity index of the recommendation system is obtained.The emergence of diversity index helps to not only pay attention to the accuracy of the recommendation,but also consider whether the recommendation information is comprehensive when evaluating the recommendation system.Experiments show that the improved Herfindahl index based on the method proposed in this experiment can accurately quantify the audience diversity of the recommendation system category.