Premium
De‐Biasing the Reporting Bias in Social Media Analytics
Author(s) -
Chen Hongyu,
Zheng Zhiqiang Eric,
Ceran Yasin
Publication year - 2016
Publication title -
production and operations management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.279
H-Index - 110
eISSN - 1937-5956
pISSN - 1059-1478
DOI - 10.1111/poms.12509
Subject(s) - social media , computer science , context (archaeology) , analytics , user generated content , weighting , social media analytics , order (exchange) , internet privacy , data science , business , world wide web , medicine , paleontology , radiology , biology , finance
User‐generated contents ( UGC ) in social media such as online reviews are inherently incomplete since we do not capture the opinions of users who do not write a review. These silent users may be systematically different than those who speak up. Such differences can be driven by users’ differing sentiments toward their shopping experiences as well as their disposition to generate UGC . Overlooking silent users’ opinions can result in a reporting bias. We develop a method to model users’ UGC generating process and then rectify this bias through an inverse probability weighting ( IPW ) approach. In the context of users’ movie review activities at Blockbuster.com, our results show that the average probability for a customer to post a review is 0.06 when the customer is unsatisfied with a movie, 0.23 when indifferent, and 0.32 when satisfied. The distribution of user's reporting probability with positive experience first‐order stochastically dominates the one with negative experience. Our approach provides a realistic solution for business managers to properly utilize incomplete UGC .