RESEARCH ARTICLE Machine learning techniques as a tool for predicting overtourism: The case of Spain José Francisco Perles-Ribes 1 | Ana Belén Ramón-Rodríguez 1 | Luis Moreno-Izquierdo 1 | María Jesús Such-Devesa 2 1 Department of Applied Economic Analysis, University of Alicante, Alicante, Spain 2 Department of Economics, University of Alcalá, Madrid, Spain Correspondence José Francisco Perles-Ribes, Faculty of Economics and Business Sciences, University of Alicante, Campus San Vicente del Raspeig, Alicante 03080, Spain. Email: jose.perles@ua.es, jfperles@gmail.com Abstract One of the most challenging tasks for tourism scientists is the prediction of potential overtourism situations in the tourist destinations. Until now, some efforts have been proposed for the purpose of establishing early warning systems. However, none of the attempts has tried to make use of a powerful prediction tool that is currently available: machine learning techniques. This article seeks to fill this gap in the existing literature by proposing the use of machine learning techniques in order to predict overtourism issues on a sample of Spanish tourist cities specialized in both, urban and sun and beach tourism products. KEYWORDS early warning system, hypothesis testing, machine learning, overtourism, prediction 1 | INTRODUCTION The tourism sector has experienced strong growth in recent years thanks to its expansion across the different layers of society and the incorporation of new countries into the international circuit. In 2018, a total of 1.4 billion tourists were recorded throughout the world. This figure was not expected to be reached until 2020 (UNWTO, 2019). At the same time, mass tourism has undergone important changes in terms of its different types and the choice of destinations. Although it began as a very rigid and standardized form of tourism based on package holidays, today we can observe greater flexibility thanks to new agents who disintermediate supply, such as low-cost airlines or collaborative accommodation (Vainikka, 2013). These transformations represent important challenges for both destination and business managers as well as researchers of the tour- ism phenomenon. Among the primary concerns is the phenomenon called overtourism, associated to the tourist overcrowding experi- enced in many destinations (Doods & Butler, 2019). This over- crowding is usually related to a lack of correct management and an uncontrolled development that surpasses its carrying capacity (UNWTO, 2018), which can lead to situations of tourismphobia or the rejection of tourists by residents (Milano, 2018; Milano, Novelli, & Cheer, 2019a). In this sense, overtourism is contemplated as the opposite scenario of the responsible and sustainable tourism develop- ment models (Jørgensen & McKercher, 2019; Peeters et al., 2018). To date, a qualitative-descriptive approach has been predominant in the debate and research on this phenomenon. From this perspec- tive, authors such as Huete & Mantecón, 2018; Koens, Postma, & Papp, 2018; Milano, 2018; Namberger, Jackisch, Schmude, & Karl, 2019; Smith, Sziva, & Olt, 2019; UNWTO, 2018) describe situa- tions of overtourism, analyze its causes and effects comparing differ- ent case studies and propose measures to mitigate the problems generated by the excessive concentration of tourists. However, significant efforts have also been made to address this problem from a quantitative point of view. These studies, with basic statistical analysis techniques (Alcalde-García, Guitart-Casalderrey, Pitarch-March, & Vallvé-Fernández, 2018; Capocchi, Vallone, Amaduzzi, & Pierotti, 2019; McKinsey&Company, 2017; Peeters et al., 2018), have sought to establish key determinants of the phe- nomenon and to set thresholds on which potential situations of over- tourism can be inferred. However, until now, it has not been possible to determine any clearly discriminating variable (overlap) on which to construct and early warning system (Peeters et al., 2018). This article seeks to shed light on the subject, attempting to establish how tourism competitiveness can become a key predictor of overtourism situations, a hypothesis established in the preliminary Received: 30 January 2020 Revised: 19 May 2020 Accepted: 22 May 2020 DOI: 10.1002/jtr.2383 Int J Tourism Res. 2020;1–14. wileyonlinelibrary.com/journal/jtr © 2020 John Wiley & Sons Ltd 1