Ajuste da série de receita por quarto (revpar) de 2016 a 2020 em um hotel no Sudeste Brasileiro, por meio do método ARIMA com análise de intervenção

Abstract

The dynamic hotel pricing strategy brings challenges to the revenue forecasting procedure. Since revenue is an essential component of the management and operation of hotel activities it is necessary to use indicators and methods that help in this task. In this scenario, there is the revenue per room indicator (revpar) which is the result of a function involving average daily rate and occupancy, two important indicators for the routines of a hotel. Therefore, from the revpar series of a hotel in southeastern Brazil collected during the period from January 2016 to December 2020, this work will perform two methods of adjustment of the series to verify which best fits: a trend and seasonality adjustment, and an ARIMA model. In addition, given the period of the series and that during the year 2020 there was an intervention of the Covid-19 pandemic, the model must be able to adjust and predict future values taking into account the presence of outliers. However, the abrupt interruption in March and April 2020 and level shift in sequence months of 2020 due to the pandemic event influenced the trend and seasonality present in the series, so the adjustment by the trend and seasonality method was not able to perform the adjustment and prediction satisfactorily. In the end, besides the expectation to fit a SARIMA model, the best fit was a non-seasonal ARIMA(1,1,1) model with intervention analysis and the non-seasonal ARIMA(1,1,1) model without intervention analysis showed the best results in terms of predictive capacity.