Modeling and Forecasting Daily Hotel Demand: A Comparison Based on SARIMAX, Neural Networks, and GARCH Models

Overnight forecasting is a crucial challenge for revenue managers because of the uncertainty associated between demand and supply. However, there limited research that focuses on predicting daily hotel demand. Hence, this paper evaluates various models’ traditional time series performances at multiple horizons. The models include seasonal naïve, Holt–Winters (HW) triple exponential smoothing, an autoregressive integrated moving average (ARIMA), (SARIMAX) with exogenous variables, multilayer perceptron (MLP) artificial neural networks model (ANNs), sGARCH, GJR-GARCH models. dataset study contains observations from in US metropolitan city 2015 to 2019 set social environmental features such as temperature, holidays, competitive ranking. Experimental results indicated under MAPE accuracy measure: (i) SARIMAX external regressors outperformed ANN-MLP similar other models, every one horizon except out seven forecast horizons; (ii) sGARCH(4, 2) GJR-GARCH(4, shows superior predictive all performance evaluated by conducting pairwise comparisons different model’s distribution forecasts using Diebold–Mariano Harvey–Leybourne–Newbold tests. are significant they provide valuable insights into variables impact accurate forecasting.