International Journal of Hospitality Management 47 (2015) 14–24 Contents lists available at ScienceDirect International Journal of Hospitality Management jou rn al hom ep age: www.elsevier.com/locate/ijhosman Hotel location evaluation: A combination of machine learning tools and web GIS Yang Yang a,1 , Jingyin Tang b,2 , Hao Luo c,∗ , Rob Law d,3 a School of Tourism and Hospitality Management, Temple University, Philadelphia, PA 19122, United States b Department of Geography, University of Florida, Gainesville, FL 32611, United States c Sun Yat-sen Business School, Sun Yat-sen University, Guangzhou, Guangdong 510275, China d School of Hotel and Tourism Management, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China a r t i c l e i n f o Keywords: Web GIS Hotel location Spatial decision making Machine learning a b s t r a c t The need for a reliable, unbiased, and objective assessment of hotel location has always been important. This study presents a new approach to evaluate potential sites for proposed hotel properties by design- ing an automated web GIS application: Hotel Location Selection and Analyzing Toolset (HoLSAT). The application uses a set of machine learning algorithms to predict various business success indicators asso- ciated with location sites. Using an example of hotel location assessment in Beijing, HoLSAT calculates and visualizes various desirable sites contingent on the specified characteristics of the proposed hotel. The approach shows considerable potential usefulness in the field of hotel location evaluation. © 2015 Elsevier Ltd. All rights reserved. 1. Introduction Evaluating and assessing a location site is an important aspect when establishing a new hotel to secure long-term business pros- perity. Once located, it is nearly impossible for the hotel to relocate; owing a considerable sunk cost for hotel establishment and oper- ation. A large body of literature has been devoted to deciphering hotel location patterns and mechanisms through different per- spectives; and several models have been advocated including mono-centric models (Egan and Nield, 2000; Yang et al., 2012), agglomeration models (Canina et al., 2005; Kalnins and Chung, 2004), and multi-dimension models (Baum and Haveman, 1997; Urtasun and Gutiérrez, 2006). Various empirical efforts have also been conducted to recognize those factors as influencing deci- sions on where to locate new hotels such as star rating, years after opening, service diversification, ownership, agglomeration effects, public service infrastructure and transport accessibility (Yang et al., 2014). In conventional hotel location prediction models, linear regres- sion has been dominantly used to recognize the preferred site ∗ Corresponding author. Tel.: +86 20 8411 2561; fax: +86 20 8411 3687. E-mail addresses: yangy@temple.edu (Y. Yang), sugar1987cn@gmail.com (J. Tang), luohao6@mail.sysu.edu.cn (H. Luo), rob.law@polyu.edu.hk (R. Law). 1 Tel.: +01 215 204 8701; fax: 01 215 204 8705. 2 Tel.: +01 352 294 7521; fax: 01 352 392 8855. 3 Tel.: +852 3400 2181; fax: +852 2362 9362. with considerable potential (Yang et al., 2014). However, the limi- tations associated with simple linear regression result in several common drawbacks such as poor prediction accuracy because of the over-fitting problem, inability to adequately incorporate interactions and nonlinearity among variables, and failure to con- sider spatial dependency and spatial heterogeneity. Furthermore, although numerous theoretical and empirical models have been constructed by scholars, the majority are fairly complicated and difficult for practitioners to understand because of the heavy use of sophisticated mathematical or statistical knowledge. Therefore, the practical value and applicability of these models become rela- tively limited, and this phenomenon impedes the wide application of these “scholarly” models to solve real hotel location problems (Yang et al., 2014). To better facilitate the decision making of hotel location selection, a huge demand has arisen for the ability to trans- fer knowledge from these “scholarly” location models to knowledge with great practical values. Over the past decade, numerous applications in information technology (IT) have been introduced in the hospitality industry to handle routine operational problems (O’Connor and Murphy, 2004; Rob and Giri, 2005). These applications have been used to improve employee productivity and increase revenue (Siguaw et al., 2000), facilitate information exchange between hotels and customers (Chung and Law, 2003), and propose appropriate room pricing across different distribution channels (O’Connor, 2003). Despite the prevalence of IT applications, few have incorporated geographical information system (GIS) technology in hospitality management. Considering that locational information is critical in http://dx.doi.org/10.1016/j.ijhm.2015.02.008 0278-4319/© 2015 Elsevier Ltd. All rights reserved.