Predicting Hotel Bookings Cancellation with a Machine Learning Classification Model

Abstract

Booking cancellations have significant impact on demand-management decisions in the hospitality industry. To mitigate the effect of cancellations, hotels implement rigid cancellation policies and overbooking tactics, which in turn can have a negative impact on revenue and on the hotel reputation. To reduce this impact, a machine learning based system prototype was developed. It makes use of the hotel's Property Management Systems data and trains a classification model every day to predict which bookings are "likely to cancel" and with that calculate net demand. This prototype, deployed in a production environment in two hotels, by enforcing A/B testing, also enables the measurement of the impact of actions taken to act upon bookings predicted as "likely to cancel". Results indicate good prototype performance and provide important indications for research progress whilst evidencing that bookings contacted by hotels cancel less than bookings not contacted.

Predictive models for hotel booking cancellation: a semi-automated analysis of the literature

“…Development or discussion of a model seems to be transversal to almost all documents since "model" is a highly frequent term in all documents except that by Metzger et al (2012). The term "hotel" shows varying frequency, which might point to the document's specificity to the hospitality industry (Antonio et al, 2017b(Antonio et al, , 2016Gayar et al, 2011;Zakhary et al, 2011). The suggestions arising from the analysis of the heatmap presented in Figure 15 are confirmed by reading the full texts of the selected documents.…”

“…The authors intended to describe a model to optimise overbooking and fare class allocation instead of building a forecasting or prediction model. The works of Antonio et al (2017bAntonio et al ( , 2017c, Gayar et al (2011), were the only ones focused on hotels while Lan et al (2011), Lemke et al (2009Lemke et al ( , 2013, Morales & Wang (2010), and Pulugurtha & Nambisan (2003) worked with airlines or employ airline data. Azadeh et al (2013), Cirillo et al (2018), and Tsai (2011) worked with railways.…”

“…Azadeh et al (2013) developed a model to predict bookings and cancellations for a railway operator. Antonio et al (2017aAntonio et al ( , 2017b employed hotel data to develop models to predict each booking's cancellation probability, and, simultaneously, net demand. Tsai (2011) used railway data to develop a model to forecast the cancellation rate.…”

“…All of the previous works employed different methods: time series-based techniques (Lemke et al, 2009(Lemke et al, , 2013, economics-based techniques (Cirillo et al, 2018;Tsai, 2011), and more modern techniques usually applied in machine learning problems, like genetic algorithms, neural networks, or other advanced classification techniques (Antonio et al, 2017b(Antonio et al, , 2017cAzadeh et al, 2013;Morales & Wang, 2010;Pulugurtha & Nambisan, 2003). Except the works by Antonio et al (2017bAntonio et al ( , 2017c and Morales & Wang (2010), who considered bookings cancellation estimation a classification problem, all other authors considered it a regression problem. The advantage of considering the problem a classification problem is that it is possible to calculate the cancellation rate (the regression measure) from the prediction of each booking´s likelihood to be cancelled (Antonio et al, 2017c).…”

“…The advantage of considering the problem a classification problem is that it is possible to calculate the cancellation rate (the regression measure) from the prediction of each booking´s likelihood to be cancelled (Antonio et al, 2017c). Although Morales & Wang (2010) argued that predicting the probability of a customer to cancel with high accuracy is difficult and that it would have no practical implications, Antonio et al (2017bAntonio et al ( , 2017c have shown the opposite. The authors showed that, in practice, the identification of bookings with a high likelihood of cancelling allows hotels to take measures to avoid the identified cancellations.…”