【摘要】 基于符号数据分析所具有的能够有效地对海量数据进行降维并从整体上把握样本属性的优势,设计了基于区间型和分布式符号数据的模型建立方法,分别建立符号数据描述的目标群体用户模型和目标项目模型,并将目标项目模型分解为积极子模型和消极子模型来表示.进而计算目标群体模型与目标项目积极子模型、消极子模型之间的相似度,最终产生推荐.选取为群体用户推荐美食作为实例,通过大众点评网收集用户评分数据,对文中算法进行评价,结果表明该算法能取得良好的推荐精度,且在群体较小及数据稀疏时,推荐质量明显优于传统基于点数据描述群体用户模型的协同过滤算法.更多还原

【Abstract】 Since symbolic data analysis can reduce the dimension of mass data and seize the sample property on the whole, a novel group user modeling method was proposed by using interval and distributed symbolic data. In particular, a symbolic data-based objective group user model and an objective item model were established, where the latter one was decomposed into an active sub-model and a passive sub-model. The similarity between the active group user model and the objective item model were calculated to generate the recommendation. Experimental results, by collecting data from the internet, illustrate that the proposed algorithm achieves satisfactory recommendation accuracy and that it can achieve superior recommendation performance over the traditional user-based point data group collaborative filtering algorithm.更多还原