A product ranking method combining the features–opinion pairs mining and interval-valued Pythagorean fuzzy sets

Abstract

Mining online reviews has become an important means of identifying consumer behavior and the innovation direction of products. However, it is difficult for both producers and consumers to effectively analyze and extract relevant opinions from a vast number of online reviews. To overcome this problem, a product ranking method that combines feature–opinion pairs mining and interval-valued Pythagorean fuzzy (IVPF) sets was proposed in this study. First, three types of important feature–opinion pairs were clearly defined based on the diversity and complexity of opinion expression forms in Chinese ecommerce reviews. Two deep learning models were then designed to automatically extract the feature–opinion terms and match them into pairs. Afterwards, sentiment analysis techniques were applied to identify sentiment orientation, and the feature–opinion pairs were clustered into groups using K-means clustering algorithm. Meanwhile, considering the confidence level based on the number of online reviews on different products, sentiment value was transformed into interval-value from, including interval membership and non-membership. As the sum of the converted interval membership and non-membership was greater than 1 and their quadratic sum was less than 1, IVPF set was introduced to represent the interval-valued sentiment. Furthermore, based on the interrelationship between product attributes, we proposed an IVPF weighted Heronian mean operator to aggregate the attribute information. Product ranking was then achieved based on the operator and operations under the IVPF information. Finally, a case study was used to verify the feasibility of the proposed method, and comparisons and sensitivity analysis were performed to demonstrate the superiority of our method.