S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Program in Technology, Management, Economics and Policy (협동과정-기술·경영·경제·정책전공) Theses (Master's Degree_협동과정-기술·경영·경제·정책전공)
A Promising Structural Importance Heuristic for Evaluating Online Reviews
- Jörn Altmann
- 공과대학 협동과정 기술경영·경제·정책전공
- Issue Date
- 서울대학교 대학원
- Social Network Analysis ; Centrality Measures ; Recommender Systems ; Online Shopping System ; Reviews ; Data Mining ; Graph Mining
- 학위논문 (석사)-- 서울대학교 대학원 : 공과대학 협동과정 기술경영·경제·정책전공, 2018. 2. Jörn Altmann.
- Representing a dramatic increase in the number of online customers and the volume of reviews, online e-commerce markets with the most successful business models continue to expand. Apart from the information offered by sellers about their products, online customer reviews are the only other information source available. Therefore, online customer reviews are a particularly useful source of information to evaluate the product quality. However, two issues arise here though: First, due to limited time, it is difficult for users to read all of the reviews. Second, the credibility of online customer reviews is frequently problematic. As demonstrated by the extant research, some sellers recognition of the power of online customer reviews for their business growth leads those sellers to recruit people to issue fake reviews on their products or services. To address this issue, in the present study, we propose a solution that combines the structural importance of reviewers and their responses towards existing reviews. Structural importance shows the importance of reviewers within social networks, enabling users to build online relationships and process information (e.g. create, read, share, re-share, comment) among them. In order to quantify structural importance of reviewers in a network, various centrality measures are used. Reviewers responses point to their reactions towards other reviewers comments. Combining these two sources of information is expected to be a promising structural importance heuristic that is based not solely on the connectedness within a network, but also considers nodes attributes. Finally, we investigate how the proposed solution can help end-users in identifying and ranking the most relevant and accurate reviews within online e-commerce platforms. The proposed solutions are planned to be run and evaluated on extensive data (i.e., millions of nodes and links and published crowd-sourced reviews) collected from Yelp.com, one of the most widely used and successful online ecommerce platforms.