Abstract:

Formal concept analysis is a data analysis method for formal context and has been introduced into the field of recommender systems. As an effective tool for formal concept analysis, concept lattice is difficult to cope with large-scale data in e-commerce because of its low construction efficiency. To solve this problem, this paper proposes a group recommendation method based on heuristic concept construction. Firstly, based on user??s common scoring items, a heuristic information is defined to speed up construction of concept. At the same time, using the intension constraint, a concept with largest area is constructed to aggregate more similar users. Then, on the concept set covering all users, group users in the concept are recommended by items popularity of the group. In the sampled data sets and MovieLens, the proposed method is compared to two different recommended methods. The experi-mental results show that the method can quickly generate a set of concepts to meet the recommendation require-ments under large-scale data.

Key words: formal concept analysis (FCA), group recommendation, heuristic algorithm, recommender system

摘要：

形式概念分析是形式背景的数据分析方法，已被引入推荐系统领域。概念格作为形式概念分析的有效工具，因其构造效率低下，所以难以应对电子商务中的大规模数据。为解决该问题，提出一种基于启发式概念构造的组推荐方法。首先，基于用户共同评分的项目，定义概念构造的启发式信息，实现概念的快速构造；同时利用内涵约束，在保证群组相似度的基础上，构造当前面积最大的概念，以包含更多的邻居用户；然后，在覆盖所有用户的概念集合上，统计项目在群组中的流行度，实现对群组用户的组推荐。在抽样数据集和MovieLens上，对比了该算法与两类不同的推荐算法。实验结果表明，在大规模数据下，该算法能在快速生成概念集合同时满足推荐需要。

关键词: 形式概念分析（FCA）, 组推荐, 启发式算法, 推荐系统