Artificial Intelligence and Smart Technologies in the Hospitality Industry

Abstract

Artificial intelligence (AI) and smart technologies are transforming the way hotels function, but most studies have focused on a single application and do not quantify its operational value end to end. In this paper, we present and test five AI modules: using a common, simulated booking dataset for a resort hotel as well as for a city hotel (2015-2017) and a strict temporal train/test split, an integrated fourlayer framework (data and sensing, AI analytics, decision and automation, value) is proposed. For booking cancellation prediction, a gradient-boosting model yields well-calibrated results with an AUC of 0.845 [95% CI 0.841–0.848], with lead time, deposit policy, and distribution channel as the most important predictors. Classical methods, for forecasting occupancy from reservations made on the books, give the best results for periods up to 1 week before arrivals; by contrast, a gradient-boosting model produces accurate results at longer lead times, about 30 days prior to arrivals at the resort (MAPE 4.4% vs. 11.8% at 90 days prior). An ML risk-based overbooking policy can improve net room revenue by 0.55% and 0.32% (resort and city, respectively) compared with the historical policy and achieve 57–59% of the value of perfect information. The "K-means Segmentation" method shows 4 segments that guests can act on. Based on an hourly smartroom simulation, IoT based HVAC control saves 23.4% energy, while AI-based HVAC predictive pre-conditioning continues to save 21.2% energy without sacrificing guest comfort. Findings reveal that the value of AI within the hospitality sector depends on the horizon as well as context and integration, and offer practitioners and researchers benchmarks and open code.