Dynamic multivariate interval forecast in tourism demand

This study proposes a dynamic multivariate interval forecasting framework for tourism demand, including variable selection, parameter optimization, and interval estimation, to simultaneously select influencing factors and their lag lengths and capture the uncertainty associated with tourism demand. The sequential association rule is used to identify key variables, while optimized support vector machines and quantile regression are applied to conduct interval forecasting. We find that both environmental factors and online search keywords are highly correlated with tourism demand. Compared to other well-known models, the proposed framework can achieve higher forecasting accuracy with lower computational complexity for tourism demand irrespective of whether it is point or interval forecasting.