Tourism demand forecasting using compound pattern recognition

Abstract

Forecasting tourism demand is challenging due to complex seasonality and unexpected crises and events. Pattern recognition has been acknowledged as an effective tool for managing this uncertainty. This study develops a compound pattern recognition framework that dynamically compounds calendar and tourism demand volume patterns to forecast daily tourism demand. Adaptive similarity evaluation and optimal combination algorithm are incorporated into this process to capture the specific characteristics in the demand. An empirical examination of three tourist attractions in China demonstrates that this novel forecasting framework has achieved sound performance during both normal periods and the crisis of COVID-19. The findings provide tourism stakeholders with an effective solution for daily tourism demand forecasting tasks.