基于偏好诱导量词的个性化推荐模型

提出一种新颖精巧的用户期望值提取模型，据此建立基于用户偏好的个性化模糊量词.首先给定一组多属性样本信息，仅要求用户根据自己的态度偏好或主观评判，提供一个关于样本方案的优劣排序.根据这个排序序列，基于有序加权平均(ordered weighted averaging, OWA)思想并利用理想解法(technique for order preference by similarity to ideal solution, TOPSIS)方法，构造用户期望值提取模型，获取用户关于样本信息的期望值，再从中抽取偏好、态度等个性特征信息，建立针对此用户的个性化量词.最后利用所得量词对新产品进行OWA数据集成，实现个性化产品推荐.案例研究及实验分析表明，所提模型及方法能够很好地捕获并反映主体的偏好及态度等个性特征，在实际应用中可面向不同层次水平、不同知识结构的用户，理性快捷地向其推荐相应态度偏好下的“最满意方案”而非一般意义下的“最优方案”，相比同类方法具有更大的实用性和灵活性.

Personalized Recommendation Model Based on Quantifier Induced by Preference

A novel model for extracting expected value from users is presented, with which to establish a user preference-based personalized quantifier. A sample of multi-attribute alternatives is given first, and then the involved user is asked to provide a ranking of alternatives of this sample on the basis of his/her personal preference or decision attitude. With this ranking, an extraction model for users’ expected value about the sample information is constructed. The principles of OWA (ordered weighted averaging) aggregations and TOPSIS (technique for order preference by similarity to ideal solution) are followed during the modeling, upon which the developed technique is based. The user’s preference or attitude can then be derived from this expected value to help build a personalized quantifier, with which to aggregate the attribute values of new products with the aim of realizing personalized recommendation. Case study and experimental results show that the developed model and quantifier can well capture and reflect many varieties of personality characteristics of users with different ability levels and knowledge structures. As such, the developed technique could be considered as an effective tool in practical applications for the “satisfactory solutions” in accord with some particular attitude, rather than the “optimal solutions” in general terms, characterized by greater applicability and flexibility by contrast with a similar kind of method.