Hotel profitability: a multilayer neural network approach

Purpose The purpose of this paper is to design an algorithm to predict hotel profitability by means of deep learning techniques. Design/methodology/approach The methodology consists of a multi-layered neural network that includes a lag of profitability as the input. Furthermore, other input variables are related to hotel and tourist destinations; the raw data for hotel and tourist destinations were collected from multiple public access data sources. Findings The results show that the proposed model has a high predictive capacity of hotel profitability in all the years studied (2005-2011), according to the performance metrics evaluated within the sample. Thus, the authors can conclude that deep learning algorithms can be a useful tool to evaluate hotel performance. Practical implications The algorithm designed in this research could be of interest to improve decision-making processes related to profitability, for example, in evaluating the creation of new hotels. Moreover, the model provides a quick and efficient analyses that could be of interest to investors and lenders. In particular, they could compare investment alternatives in the hotel sector. Also, according to the results, the location variables are important determinants of hotel profitability, and consequently, hotel managers should collaborate with the tourist destination managers to improve profitability. From an internal perspective, hotel managers should focus on the management of human resources. Originality/value This paper is the first empirical study that predicts hotel profitability using deep learning techniques. In addition, this methodology is applied to analyse hotel profitability, for the first time, in the Spanish market. This market is an ideal analytical framework because of its heterogeneity with respect to hotel supply in terms of seasonality and coastal characteristics, among others.