Abstract: In recent years, with the rapid development of computer technology and e-commerce, the online reviews given by consumers or the public for the products can be seen on various websites, and the online reviews have a certain guiding role in the purchase decision of potential consumers. At present, consumers evaluate products or programs from multiple attributes, such as service level, logistics level, product quality, function, price and so on. Consumers use words such as “excellent, good, medium, and poor”, to express the evaluation of attributes of products or programs. As we all known, attributes are not absolutely independent of each other, and there is an association relationship between attributes actually. The online reviews on the website come from different reviewers or consumers, and the reliability of online reviews varies due to the different knowledge, experience and preferences of reviewers. Therefore, in the process of analysis and decision making, how to determine the recommendation ranking among the alternatives according to the large amount of online reviews on the website considering the differences in the reliability of online reviews with attributes association, is an important and significant research topic.
   To solve the problem of multi-attribute decision making considering the reliability of online reviews with attribute association, a multi-attribute decision making method based on the 2-tuple linguistic Choquet integral operator of online reviews is proposed. Firstly, consumers’ online reviews are mainly stored on various websites in the form of natural language. On the one hand, the use of 2-tuple linguistic to describe online reviews can effectively avoid information loss and distortion in the process of aggregation and operation. On the other hand, the deviation of linguistic is considered to make the calculation result more precise. So the 2-tuple decision matrix is constructed to describe online reviews by using 2-tuple. Secondly, the reliability of online reviews varies due to the knowledge, experience, and preferences of reviewers. In the process of evaluating customer satisfaction, if we ignore the reliability difference of online reviews and treat different online reviews equally, the accuracy of the evaluation results will be reduced. In view of this, this paper proposes a method on reliability of online reviews which considers the release time of online reviews, the number of “useful” votes obtained, the rating of reviewers, praise rewards, limited rational behavior of decision makers and so on. Thirdly, in the multi-attribute decision-making environment, it is difficult to maintain absolute independence among attributes, and the attribute association may have an impact on the decision result. Therefore, for the multi-attribute decision-making problem with attribute association, the determination of attribute weight should consider its contribution to the decision result, and the definition of Shapley value is used to determine the attribute weight in this paper. On this basis, the 2-tuple Choquet integral operator is used to aggregate the decision matrix to obtain the alternatives’ comprehensive values and ranked accordingly. Finally, in order to verify the rationality and feasibility of the proposed method in this paper, four hotels with similar orientation and room charges on the website “www.qunar.com” are taken as example, and the alternatives ranking obtained by using the proposed method is consistent with the rating ranking of each hotel on the website.
   In future work, several directions can extend our study. The online reviews in the form of graphics, videos and texts, and the influence of malicious negative reviews on decision making can be considered in the mathematical model.

Key words: online reviews, 2-tuple linguistic Choquet integral operator, reliability, attribute association, multi-attribute decision making

摘要： 考虑在线评价信息的可信度及决策指标的关联性,提出了一种基于在线评价信息的二元语义Choquet积分算子的多属性决策方法。首先,为了有效避免语言评价信息集结和运算中出现的信息损失和扭曲,利用二元语义描述在线评价信息,构建二元语义决策矩阵。其次,考虑在线评价信息的发布时间、获得的“有用”投票数、评论者等级、好评奖励等因素,结合决策者的有限理性行为,确定在线评价信息的可信度。再次,针对属性间具有相互作用关系的多属性决策问题,属性权重的确定应考虑其对决策结果的贡献程度,所以参考Shapley值的定义确定属性权重。在此基础上,利用二元语义Choquet积分算子对决策矩阵进行集结求得方案综合值。最后,通过一个实例说明所提方法的合理性及可行性。

关键词: 在线评价信息, 二元语义Choquet积分算子, 可信度, 属性相关, 多属性决策