Decision analysis framework based on incomplete online textual reviews

Abstract

Information is a key factor that influences the performance of decision makers. With the explosive proliferation of Web 2.0, the volume of online textual reviews has been sharply increasing. However, how to use this type of unstructured data and utilize the valuable information hidden behind are still problems to be solved. This study aims to provide a multi-attributes decision analysis (MADA) framework based on incomplete online textual reviews to aid in decision making. First, online textual reviews are obtained by data crawling. Attributes information is determined by textual analysis and the attitudes of assessors toward each attribute is discriminated by the sentiment analysis. Then, some new rules are developed in encoding incomplete online textual reviews into interval-valued linguistic distribution assessment (ILDA) to better characterize the evaluators’ attitudes. Next, evidential reasoning (ER) algorithm is extended to the ILDA environment to combine the information with multiple attributes, and the utility interval of each alternative is constructed by solving a pair of nonlinear optimization models. Given that the interval data cannot be directly compared, an enhanced minimax regret approach is proposed to compare and rank them. Finally, a real case study about online commodity evaluation is examined to show the implementation process of the proposed framework, and a discussion is also conducted to systematically analyze its superiority.