Predicting Airbnb pricing: a comparative analysis of artificial intelligence and traditional approaches

Abstract

Airbnb's distinctive model accommodates a broad spectrum of hosts, ranging from non-professionals to traditional establishments, resulting in a nuanced pricing system that poses challenges for prediction. This paper delves into Airbnb pricing and, to address this exercise, introduces computational approaches that combine traditional linear methods and advanced artificial intelligence techniques. Utilizing real data from the Netherlands, the study specifically focuses on all active Airbnb listings until September 2019. Our contribution stands out by incorporating the financial history of diverse rental offerings, a novel aspect compared to existing literature. Unlike other studies, our investigation spans various cities in the Netherlands, providing a comprehensive view beyond individual regions. Our results highlight the robust predictive capabilities of Artificial Intelligence techniques, that lead to lower susceptibility to overfitting and to superior overall performances. However, traditional methods also prove to be valuable, aiding in determining the significance of predictors in the predicting exercise. Collectively, these approaches offer valuable insights to enhance pricing strategies and overall performances within the Airbnb market.