Abstract: With the rapid development of e-commerce and social network, the number of online goods and services is increasing day by day, making it more and more difficult for customers to make purchase decisions based on online information. A product recommendation model is constructed to help customers successfully locate the products customers need based on online numerical ratings with large data volume. Firstly, an intuitionistic trapezoidal fuzzy number is introduced to convert the numerical ratings to deal with the characteristics of uncertainty in historical users' ratings. Secondly, the classical prospect theory is introduced to depict the loss aversion of customers in the realistic decision-making process. At last, the practicability and effectiveness of the model are verified through the analysis of the automobile purchase cases of the Autohome and the comparison with the recommendation result of completely rational customers.

Key words: product recommendation, numerical rating, intuitionistic trapezoidal fuzzy number, prospect theory

摘要： 随着电子商务和社交网络的快速发展,在线商品和服务的数量日益增长,顾客越来越难以依据在线信息制定其购买决策。本文基于数据体量较大的在线数值评分,构建产品推荐模型,帮助顾客成功定位所需商品。首先,引入直觉梯形模糊数进行数值评分转换,以处理历史用户给定评分数据的不确定性特征;其次,引入经典前景理论来刻画现实决策过程中顾客的损失厌恶心理。最后,通过汽车之家网站的汽车购买案例及与完全理性顾客推荐结果的对比分析,验证了该模型的实用性和有效性。

关键词: 产品推荐, 数值评分, 直觉梯形模糊数, 前景理论