Multi-attraction, hourly tourism demand forecasting

Zheng, Weimin; Huang, Liyao; Lin, Zhibin

doi:10.1016/j.annals.2021.103271

Cited by 38 publications

(21 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Vector autoregressive (VAR) models, such as Assaf et al (2018), Gunter and Önder (2016) and Torraleja et al (2009), and error corrections and autoregressive distributed lag are widely used as well. AI-based methods, including neural networks (NNs) such as Claveria et al (2015, 2017), Höpken et al (2020), Hu et al (2019), Silva et al (2019) and Xie et al (2020), deep learning such as Bi et al (2021), Kulshrestha et al (2020) and Zheng et al (2021), long short-term memory such as Bi et al (2020) and Wu et al (2021), and support vector regression (SVR) such as Chen (2011), Chen and Wang (2007) and Fang et al (2021), not only do not require any statistical assumptions, but also their strong feasibility and flexibility for nonlinear data have been clearly demonstrated (Bi et al , 2020).…”

Section: Literature Reviewmentioning

confidence: 99%

The impact of Google Trends index and encompassing tests on forecast combinations in tourism

2022

View full text Add to dashboard Cite

Purpose Given that the use of Google Trends data is helpful to improve forecasting performance, this study aims to investigate whether the precision of forecast combination can benefit from the use of Google Trends Web search index along with the encompassing set. Design/methodology/approach Grey prediction models generate single-model forecasts, while Google Trends index serves as an explanatory variable for multivariate models. Then, three combination sets, including sets of univariate models (CUGM), all constituents (CAGM) and constituents that survive the forecast encompassing tests (CSET), are generated. Finally, commonly used combination methods combine the individual forecasts for each combination set. Findings The tourism volumes of four frequently searched-for cities in Taiwan are used to evaluate the accuracy of three combination sets. The encompassing tests show that multivariate grey models play a role to be reckoned with in forecast combinations. Furthermore, the empirical results indicate the usefulness of Google Trends index and encompassing tests for linear combination methods because linear combination methods coupled with CSET outperformed that coupled with CAGM and CUGM. Practical implications With Google Trends Web search index, the tourism sector may benefit from the use of linear combinations of constituents that survive encompassing tests to formulate business strategies for tourist destinations. A good forecasting practice by estimating ex ante forecasts post-COVID-19 can be further provided by scenario forecasting. Originality/value To improve the accuracy of combination forecasting, this research verifies the correlation between Google Trends index and combined forecasts in tourism along with encompassing tests.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

The impact of Google Trends index and encompassing tests on forecast combinations in tourism

2022

View full text Add to dashboard Cite

show abstract

“…Second, the conventional time series components embedded in these models may be unable to represent the complex time‐varying patterns of tourism demand, resulting in inaccurate hourly forecasting. Zheng et al (2021) recently proposed an AI model that addresses the second limitation by learning complex temporal patterns using AI techniques while accounting for spatial dependency simultaneously. However, they were unable to address the first limitation because they used a spatial weight matrix to represent the interactions.…”

Section: Literature Reviewmentioning

confidence: 99%

“…Existing time series and econometric models focus on forecasting demands of relatively long time spans and typically have fixed structures with global settings, thereby rendering them ineffective in dealing with nonstationary and other complex temporal patterns of tourism data commonly observed in relatively short time spans (Song & Li, 2008). Previous studies have also revealed some disadvantages of the econometric models (Zheng et al, 2021). First, involving influential factors in the model means that the factors themselves have to be forecasted.…”

Section: Models For Forecasting Tourism Demandmentioning

confidence: 99%

“…However, as mentioned above, these traditional models with fixed structures cannot handle nonlinear patterns in timely collected data. The recent study by Zheng et al (2021) addressed this limitation by leveraging artificial intelligence (AI) techniques while incorporating spatial dependencies. However, their proposed model requires a spatial weight matrix that represents the interaction structure among destinations.…”

Section: Introductionmentioning

confidence: 99%

“…Most existing studies have focused on relatively long time spans (e.g., monthly or annually) of a relatively large area (e.g., country), and their forecasts are valuable for developing macro and long-term policies as well as planning tourism infrastructures (Song et al, 2019). However, studies on forecasting hourly tourist arrivals simultaneously at multiple attractions in the same region have been nonexistent until recently (Zheng et al, 2021). The present study aims to forecast hourly tourist arrivals at multiple attractions while explicitly considering the interactions among these attractions in a data-driven manner and investigating their roles in hourly forecasts.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Data‐driven inference of interactions among multiple tourist attractions for hourly demand forecasting

Rong

Wang

Zheng

2022

Journal of Tourism Research

Self Cite

View full text Add to dashboard Cite

Existing tourism demand forecasting models mainly focus on forecasting demands of relatively long time spans at a single destination. These studies lack considering the evolution of demand patterns or presume fixed interaction structures among multiple destinations, limiting their applications during uncertain times when demands and their interactions could evolve fast. This study simultaneously infers interactions among multiple tourist attractions from tourist arrival data and learns temporal patterns for forecasting short-term tourist arrivals at these attractions. We achieve this goal by working on data of high spatiotemporal resolution and developing a variational autoencoder framework, where a data-driven encoder infers the interactions and a decoder learns the short-term dynamics of tourism demand.

show abstract

Design of a bi-level PSO based modular neural network for multi-step time series prediction

Li,

Liu,

Chen

2024

Appl Intell

View full text Add to dashboard Cite

Multi-attraction, hourly tourism demand forecasting

Cited by 38 publications

References 55 publications

The impact of Google Trends index and encompassing tests on forecast combinations in tourism

The impact of Google Trends index and encompassing tests on forecast combinations in tourism

Data‐driven inference of interactions among multiple tourist attractions for hourly demand forecasting

Design of a bi-level PSO based modular neural network for multi-step time series prediction

Contact Info

Product

Resources

About