Examining Cases of Property Misinformation in Airbnb
Restricted (Penn State Only)
Author:
Devarakonda, Vaishali
Area of Honors:
Data Sciences
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
J. Andrew Petersen, Thesis Supervisor John Yen, Thesis Honors Advisor Mallory Marie Meehan, Faculty Reader
Keywords:
Airbnb Machine Learning Misinformation Text Classification Sharing Economy
Abstract:
Sharing economies are defined by their peer-to-peer approach to conducting essential business activities and are often facilitated through a platform. These platforms are also well known to depend on consumer product reviews to drive their operations. Airbnb, an online platform for lodgings and rentals, follows this model. Those who put up their spaces online are also able to rent others' properties for a given amount of time. Most reviews for Airbnb listings are positive, but there can be a discrepancy between customer ratings and listing descriptions that results in guests alleging property misinformation. While this phenomenon is rare, it can have significant downstream consequences for Airbnb and hosts who market on the platform. Thus, I look to create a model that determines when cases of property misinformation occur on the platform and assess the implications of this phenomenon for Airbnb. Overall, classes of misinformation are most accurately predicted with an ensemble of BERT-based classifiers, and the downstream consequences of misinformation are significantly negative in terms of revenue generation for Airbnb and hosts with property misinformation in their descriptions.
