传统群组推荐算法基于点数据描述群组用户模型,存在着信息缺失、很难统筹考虑所有个体用户的需求等问题.针对该问题,对个体评分数据按照符号数据分析的思想进行“打包”,将群组成员的评分信息汇总为区间型符号数据.在Hausdorff距离基础上,采用区间内部点数据的描述统计量,提出了一种全新的区间数距离度量方法,并利用这种距离对区间型符号数据描述的群组实施K-均值聚类,由此确定相似群组,最后通过最近邻的评分预测目标群组的评分.将这种全新的群组推荐算法与传统方法进行推荐精度与效率的对比实验,结果表明,在各种实验条件下,基于区间型符号数据的群组推荐算法均优于传统点数据的群组推荐算法.%The group user profile in traditional group recommendation is described by single-valued data. This results in the loss of data information and being difficult to meet the demands of all the memebers of the group. Aimed at this problem, this paper took the method of symbolic data analysis aggregating individual ratings of the group into interval symbolic data into account. It proposed a novel distance considering the descriptive statistics of individuals within the intervals. Based on the K-means clustering on the interval data of group ratings, it obtained the similar groups. Then it predicted the ratings of the target group by using the neighbors' ratings. It conducted a simulation study to evaluate the new method. The result shows that the new method based on interval symbolic data analysis is more accurate and efficient than the traditional item-based collaborative filtering algorithms for group recommendation.
展开▼
