Abstract: Airbnb is one of the largest accommodation rental platforms. This paper comprehensively uses a variety of machine learning methods to predict Airbnb price in Beijing and examine the impacts of six groups of explanatory variables: characteristics of listings, location of listings, amenities and services, rentalrules, attributes of hosts,and reputation of listings. Findings find that: (a)Characteristics of listings have the greatest impact on the Airbnb price. (b)The performance of the nonlinear method is significantly better than the linear method. (c)Contrary to previous findings, instant booking and high price are related. (d)Among all variables, capacity of listings, room type, and the distance between listings and the city center are the most important variables. This paper provides new perspectives for the study of Airbnb pricing, and can help hosts to set price for their listings more accurately. The authors have made possible explanations from the theoretical and practical perspective.

Key words: sharing economy, airbnb, machine learning, variable selection, prediction, price determinants

摘要： Airbnb是全球最大的旅游房屋租赁平台之一。本文综合利用多种机器学习方法,基于房源本身的特征、房源的位置、设施与服务、租赁规则、房东的特征和房源的声誉这六类解释变量对北京市Airbnb房源价格进行了预测,并探讨了六类解释变量对房源价格的影响。发现:(a)六类变量中房源本身的特征对房源价格的影响最大。(b)非线性方法的表现明显优于线性方法。(c)与以往研究发现的Airbnb房源价格与允许即时预定呈负相关关系相反,北京市允许即时预订的房源价格高于不允许即时预订的房源。(d)房源容量、房间类型以及房源与市中心的距离是最重要的影响因素。本文为Airbnb房源定价研究提供了新的视角,同时有利于Airbnb房东更好地为其房源设置价格,作者从理论和实践的角度作出了可能的解释。

关键词: 共享经济, Airbnb, 机器学习, 变量选择, 预测, 价格影响因素