No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Tourism demand modelling and forecasting: a Horizon 2050 paper
Abstract
Purpose
The aim of this paper is to provide a narrative review of previous research on tourism demand modelling and forecasting and potential future developments.
Design/methodology/approach
A narrative approach is taken in this review of the current body of knowledge.
Findings
Significant methodological advancements in tourism demand modelling and forecasting over the past two decades are identified.
Originality/value
The distinct characteristics of the various methods applied in the field are summarised and a research agenda for future investigations is proposed.
... AI is a technological tool that helps create individualised offers that satisfy tourists' wants and needs, providing hyper-personalised information within specific and ubiquitous contexts [39]. Tourism demand is highly related to tourism marketing and AI-based models to encourage tourists to visit cities and their tourist attractions [40,41]. Tourism marketing is an activity that has the ability to contribute to commercial and social-economic growth in cities, as well as enhance tourists' experiences and communication with customers [42]. ...
Artificial intelligence (AI) is a disruptive technology that is being used by smart tourist destinations (STDs) to develop new business models and marketing services to increase tourists' experiences and sales, revenue, productivity, and efficiency and STDs. However, the adoption of AI applications and platforms requires a high economic budget for STDs that want to integrate this digital tool into their future agenda and tourism development plans, especially when they set them up for marketing plans and operational processes. This iterative technology needs regular maintenance as well, leading to recurring costs and specialised crews in advanced technologies and marketing activities. This study aims to show the impact of AI advancements on STDs' tourism marketing to enhance the quality of services and illustrate their future agenda to improve tourists' experiences. A comprehensive literature review on AI technology and STDs has been conducted to illustrate new tourism marketing in their future agenda. Moreover, this study presents real examples of AI technology in a tourism context to better understand the potential of this digital tool. The findings of the current study support the idea that AI is a multipurpose tool that helps manage, monitor, and analyse sales information; revenue management; minimise prediction errors; streamline operations; and develop better marketing strategies, optimising economic resources, reducing marketing costs, and responding dynamically to changing needs for tourists and residents in STDs. Furthermore, the investment in AI technologies by STDs helps enhance the quality of products and services, and attract new investments, which benefit the regional economies and population's quality of life. This study is the first to address the use of AI to improve tourist marketing in STDs, which is its primary uniqueness. Also, this study identifies new opportunities and initiatives through AI that can be developed to help tourism marketing in STDs.
Effective tourist demand forecasting is crucial for company operations and destination management. Furthermore, tourists may plan better personalized multi-attraction itineraries based on demand forecasting to avoid travel peaks and improve the enjoyment of their vacation. This study developed a unique deep learning model called the convolution block attention module (CBAM) that is built on convolutional blocks and attention modules to estimate tourism demand precisely. Then, the passenger flow grid map was extracted from mobile phone signaling data. To forecast the subsequent period of the passenger flow grid map, the CBAM model uses the multi-channel spatial-temporal grid graph that is built by multiple successive passenger flow grid maps. Finally, the forecasted passenger flow grid map was used to derive the tourist demand for multi-attractions for the next period. The analysis of mobile phone signaling data from Beijing and Xiamen using the proposed model reveals that its mean absolute percentage error (MAPE) is 8.11%, which is lower than other benchmark deep learning models.
Purpose
The rapid growth of artificial intelligence is disrupting various industries, including the tourism sector. This paper aims to outline the applications, benefits and risks of ChatGPT and large language models in general on tourism. It also aims to establish a research agenda for investigating the implications of these models in tourism.
Design/methodology/approach
Drawing on the available literature on ChatGPT, large language models and artificial intelligence, the paper identifies areas of application of ChatGPT for several tourism stakeholders. Potential benefits and risks are then considered.
Findings
ChatGPT and other similar models are likely to have a profound impact on several tourism processes. They will contribute to further streamline customer service in front-of-house operations and increase productivity and efficiency in back-of-house operations. Although negative consequences for human resources are expected, this technology mostly enhances tourism employees.
Originality/value
To the best of the authors’ knowledge, this is one of the first studies that explore the potential implications of ChatGPT in tourism and hospitality.
Forecasting tourism volume can provide helpful information support for decision-making in managing tourist attractions. However, existing studies have focused on the long-term and large-scale prediction and scarcely considered high-frequency and micro-scale ones. In addition, the current approaches are limited regarding forecasting the visitor volume of a designated sub-area in a tourist attraction. This sub-area forecast can assist local-scaled managing decisions of tourist attractions, particularly for large-scale tourist attractions. Therefore, to achieve high-frequency forecasts of tourist volume for finer scale areas such as parks and their sub-areas and generate more controllable and flexible forecasts, this study developed a novel method that incorporates a forecasting model composed of multiple deep learning components and a designed control mechanism. The control mechanism produces high-temporal-resolution sequences of tourist volume for designated sub-areas, and the forecasting model is built on an attention-based deep-bidirectional neural network to better capture the long-range dependencies of the sequence and enhance the forecasting accuracy and robustness. The experimental research was performed at Taiyangdao Park and its two designated sub-areas to validate the effectiveness and superiority of the proposed method compared to other widely used deep-learning methods; three types of performance evaluations were adopted including fitting methods, error measures, and Diebold-Mariano tests. The results demonstrated that the proposed method provided outstanding performance in high-frequency forecasts and yielded more desired forecasting outcomes than other widely used forecasting methods. Furthermore, the comparison with the performances of various other deep learning models provide insights concerning their forecasting capacity; for instance, bidirectional RNN models tend to achieve better forecasts than general RNN models in the high-frequency forecasts. The proposed method has significant practical applicability in aiding short-term micro-scale management decisions and can also serve as an alternative approach in the field of tourist volume forecasting.
Hot-spring tourism refers to entertainment, health preservation, commercial conferences, and other leisure activities at hot-spring locations. This tourism often shows periodic variability, which may be due to seasonal temperature variations. However, studies on the effects of temperature variations on tourist arrivals at hot springs are limited. Therefore, this study aimed to evaluate this relationship in 31 provincial capital cities and 13 s-tier cities in China. Using the Baidu Index, we obtained data for tourist arrivals to hot springs in each city and constructed a generalised additive model to explore the associations between temperature variations and tourist arrivals. We also analysed the statistical significance of the estimated effects during different seasons to explore potential effect modification. A 1 °C increase in temperature was associated with a 1.81% (95% confidence interval (CI): 1.69–1.93) decrease in daily tourist arrivals for hot-spring tourism. Significant positive associations between the abovementioned factors were observed in summer (2.18% change, 95% CI: 1.32–3.04). The effect of temperature on the volume of tourist arrivals may last for approximately 2 months. Robustness analysis confirmed the data reliability. The results indicate that significant relationships exist between temperature variations and hot-spring tourism arrivals, which vary seasonally. This study has significant implications for travel agencies to effectively manage tourist visits to hot spring locations.
To explore recent progress in tourism demand research, we comprehensively survey current studies in the leading tourism and hospitality journals, asking six evaluative questions about the scientific merits of the studies and three explorative questions about emerging areas in the literature. The examination identifies potential flaws and their consequences in the field of tourism demand. A theoretical foundation is recommended for future tourism demand studies with a view to reduce bias in the empirical analysis of tourism demand. Several emerging areas of analysis in the field of tourism demand are recognized and discussed. Our study provides critical insights that will enable future tourism demand research to generate more reliable, impactful information than in the past.
Purchases of voluntary carbon offsets (VCOs) are growing tremendously. At the same time, the number of activities and products for which VCOs are available is increasing. Experts discuss whether offering VCOs is exclusively associated with positive effects on the environment or if it instead may lead to increased consumption of environmentally critical products. To date, empirical evidence on such adverse effects is scarce. Therefore, this study uses a randomized controlled trial design to investigate how the availability of VCOs affects consumers' choices for environmentally critical products. The results suggest that when VCOs are available, the likelihood of environmentally critical consumption increases. From a mental accounting perspective, our findings support the theoretical rationale of VCOs as an instrument for moral licensing. Additionally, our results indicate that individuals tend to trivialize the harmfulness of the environmentally critical product and overestimate the effectiveness of VCOs for environmental protection, which we consider strategies for reducing cognitive dissonance and guilt.
COVID-19 disrupted international tourism worldwide, subsequently presenting forecasters with a challenging conundrum. In this competition, we predict international arrivals for 20 destinations in two phases: (i) Ex post forecasts pre-COVID; (ii) Ex ante forecasts during and after the pandemic up to end 2021. Our results show that univariate combined with cross-sectional hierarchical forecasting techniques (THieF-ETS) outperform multivariate models pre-COVID. Scenarios were developed based on judgemental adjustment of the THieF-ETS baseline forecasts. Analysts provided a regional view on the most likely path to normal, based on country-specific regulations, macroeconomic conditions, seasonal factors and vaccine development. Results show an average recovery of 58% compared to 2019 tourist arrivals in the 20 destinations under the medium scenario; severe, it is 34% and mild, 80%.
Purpose
This paper aims to illustrate the potential of high-frequency data for tourism and hospitality analysis, through two research objectives: First, this study describes and test a novel high-frequency forecasting methodology applied on big data characterized by fine-grained time and spatial resolution; Second, this paper elaborates on those estimates’ usefulness for visitors and tourism public and private stakeholders, whose decisions are increasingly focusing on short-time horizons.
Design/methodology/approach
This study uses the technical communications between mobile devices and WiFi networks to build a high frequency and precise geolocation of big data. The empirical section compares the forecasting accuracy of several artificial intelligence and time series models.
Findings
The results robustly indicate the long short-term memory networks model superiority, both for in-sample and out-of-sample forecasting. Hence, the proposed methodology provides estimates which are remarkably better than making short-time decision considering the current number of residents and visitors (Naïve I model).
Practical implications
A discussion section exemplifies how high-frequency forecasts can be incorporated into tourism information and management tools to improve visitors’ experience and tourism stakeholders’ decision-making. Particularly, the paper details its applicability to managing overtourism and Covid-19 mitigating measures.
Originality/value
High-frequency forecast is new in tourism studies and the discussion sheds light on the relevance of this time horizon for dealing with some current tourism challenges. For many tourism-related issues, what to do next is not anymore what to do tomorrow or the next week.
Plain Language Summary
This research initiates high-frequency forecasting in tourism and hospitality studies. Additionally, we detail several examples of how anticipating urban crowdedness requires high-frequency data and can improve visitors’ experience and public and private decision-making.
It is important to provide scientific assessments concerning the future of tourism under the uncertainty surrounding COVID-19. To this purpose, this paper presents a two-stage three-scenario forecast framework for inbound-tourism demand across 20 countries. The main findings are as follows: in the first-stage ex-post forecasts, the stacking models are more accurate and robust, especially when combining five single models. The second-stage ex-ante forecasts are based on three recovery scenarios: a mild case assuming a V-shaped recovery, a medium one with a V/U-shaped, and a severe one with an L-shaped. The forecast results show a wide range of recovery (10%-70%) in 2021 compared to 2019. This two-stage three-scenario framework contributes to the improvement in the accuracy and robustness of tourism demand forecasting.
The profound impact of the coronavirus disease 2019 (COVID-19) pandemic on global tourism activity has rendered forecasts of tourism demand obsolete. Accordingly, scholars have begun to seek the best methods to predict the recovery of tourism from the devastating effects of COVID-19. In this study, econometric and judgmental methods were combined to forecast the possible paths to tourism recovery in Hong Kong. The autoregressive distributed lag-error correction model was used to generate baseline forecasts, and Delphi adjustments based on different recovery scenarios were performed to reflect different levels of severity in terms of the pandemic's influence. These forecasts were also used to evaluate the economic effects of the COVID-19 pandemic on the tourism industry in Hong Kong.
This study proposes a hierarchical pattern recognition method for tourism demand forecasting. The hierarchy consists of three tiers: the first tier recognizes the calendar pattern of tourism demand, identifying work days and holidays and integrating “floating holidays.” The second tier recognizes the tourism demand pattern in the data stream for different calendar pattern groups. The third tier generates forecasts of future tourism demand. Evidence from daily tourist visits to three attractions in China shows that the proposed method is effective in forecasting daily tourism demand. Moreover, the treatment of “floating holidays” turns out to be more effective and flexible than the commonly adopted dummy variable approach.
This paper investigates the expected results of the current COVID-19 outbreak to arrivals of Chinese tourists to the USA and Australia. The growing market share of Chinese tourism and the fact that the county was the first to experience the pandemic make China a suitable proxy for predictions on global tourism. We employ data from the 2003 SARS outbreak to train a deep learning artificial neural network named Long Short Term Memory (LSTM). The neural network is calibrated for the particulars of the current pandemic. Our findings, which are cross-validated using backtesting, suggest that recovery of arrivals to pre-crisis levels can take from 6 to 12 months and this can have significant adverse effects not only on the tourism industry but also on other sectors that interact with it. JEL classification: H12, P46, Z32
This study investigates whether tourism forecasting accuracy is improved by incorporating spatial dependence and spatial heterogeneity. One- to three-step-ahead forecasts of tourist arrivals were generated using global and local spatiotemporal autoregressive models for 37 European countries and the forecasting performance was compared with that of benchmark models including autoregressive moving average, exponential smoothing and Naïve 1 models. For all forecasting horizons, the two spatial models outperformed the non-spatial models. The superior forecasting performance of the local model suggests that the full reflection of spatial heterogeneity can improve the accuracy of tourism forecasting.
Decomposing complex time series into trend, seasonality, and remainder components is an important task to facilitate time series anomaly detection and forecasting. Although numerous methods have been proposed, there are still many time series characteristics exhibiting in real-world data which are not addressed properly, including 1) ability to handle seasonality fluctuation and shift, and abrupt change in trend and reminder; 2) robustness on data with anomalies; 3) applicability on time series with long seasonality period. In the paper, we propose a novel and generic time series decomposition algorithm to address these challenges. Specifically, we extract the trend component robustly by solving a regression problem using the least absolute deviations loss with sparse regularization. Based on the extracted trend, we apply the the non-local seasonal filtering to extract the seasonality component. This process is repeated until accurate decomposition is obtained. Experiments on different synthetic and real-world time series datasets demonstrate that our method outperforms existing solutions.
This study reviews 211 key papers published between 1968 and 2018, for a better understanding of how the methods of tourism demand forecasting have evolved over time. The key findings, drawn from comparisons of method-performance profiles over time, are that forecasting models have grown more diversified, that these models have been combined, and that the accuracy of forecasting has been improved. Given the complexity of determining tourism demand, there is no single method that performs well for all situations, and the evolution of forecasting methods is still ongoing.
This article also launches the Annals of Tourism Research Curated Collection on tourism demand forecasting, which contains past and hot off the press work on the topic and will continue to grow as new articles on the topic appear in Annals.
Combination is an effective way to improve tourism forecasting accuracy. However, empirical evidence is limited to point forecasts. Given that interval forecasts can provide more comprehensive information, it is important to consider both point and interval forecasts for decision-making. Using Hong Kong tourism demand as an empirical case, this study is the first to examine if and how the combination can improve interval forecasting accuracy for tourism demand. Winkler scores are employed to measure interval forecasting performance. Empirical results show that combination improves the accuracy of tourism interval forecasting for different forecasting horizons. The findings provide government and industry practitioners with guidelines for producing accurate interval forecasts that benefit their policy-making for a wide array of applications in practice.
This article also launches the Annals of Tourism Research Curated Collection on Tourism Demand Forecast, a special selection of research in this field.
This study conducts spatial-temporal forecasting to predict inbound tourism demand in 29 Chinese provincial regions. Eight models are estimated among a-spatial models (autoregressive integrated moving average [ARIMA] model and unobserved component model [UCM]) and spatial-temporal models (dynamic spatial panel models and space-time autoregressive moving average [STARMA] models with different specifications of spatial weighting matrices). An ex-ante forecasting exercise is conducted with these models to compare their one-/two-step-ahead predictions. The results indicate that spatial-temporal forecasting outperforms the a-spatial counterpart in terms of average forecasting error. Auxiliary regression finds the relative error of spatial-temporal forecasting to be lower in regions characterized by a stronger level of local spatial association. Lastly, theoretical and practical implications are provided.
This study investigates whether pooling can improve the forecasting performance of tourism demand models. The short-term domestic tourism demand forecasts for 341 cities in China using panel data (pooled) models are compared with individual ordinary least squares (OLS) and naïve benchmark models. The pooled OLS model demonstrates much worse forecasting performance than the other models. This indicates the huge heterogeneity of tourism across cities in China. A marked improvement with the inclusion of fixed effects suggests that destination features that stay the same or vary very little over time can explain most of the heterogeneity. Adding spatial effects to the panel data models also increases forecasting accuracy, although the improvement is small. The spatial distribution of spillover effects is drawn on a map and a spatial pattern is recognized. Finally, when both spatial and temporal effects are taken into account, pooling improves forecasting performance.
Increasing levels of global and regional integration have led to tourist flows between countries becoming closely linked. These links should be considered when modeling and forecasting international tourism demand within a region. This study introduces a comprehensive and accurate systematic approach to tourism demand analysis, based on a Bayesian global vector autoregressive (BGVAR) model. An empirical study of international tourist flows in nine countries in Southeast Asia demonstrates the ability of the BGVAR model to capture the spillover effects of international tourism demand in this region. The study provides clear evidence that the BGVAR model consistently outperforms three other alternative VAR model versions throughout one- to four-quarters-ahead forecasting horizons. The potential of the BGVAR model in future applications is demonstrated by its superiority in both modeling and forecasting tourism demand.
Purpose
The purpose of this paper is to review recent studies published from 2007 to 2015 on tourism and hotel demand modeling and forecasting with a view to identifying the emerging topics and methods studied and to pointing future research directions in the field.
Design/methodology/approach
Articles on tourism and hotel demand modeling and forecasting published mostly in both science citation index and social sciences citation index journals were identified and analyzed.
Findings
This review finds that the studies focused on hotel demand are relatively less than those on tourism demand. It is also observed that more and more studies have moved away from the aggregate tourism demand analysis, whereas disaggregate markets and niche products have attracted increasing attention. Some studies have gone beyond neoclassical economic theory to seek additional explanations of the dynamics of tourism and hotel demand, such as environmental factors, tourist online behavior and consumer confidence indicators, among others. More sophisticated techniques such as nonlinear smooth transition regression, mixed-frequency modeling technique and nonparametric singular spectrum analysis have also been introduced to this research area.
Research limitations/implications
The main limitation of this review is that the articles included in this study only cover the English literature. Future review of this kind should also include articles published in other languages. The review provides a useful guide for researchers who are interested in future research on tourism and hotel demand modeling and forecasting.
Practical implications
This review provides important suggestions and recommendations for improving the efficiency of tourism and hospitality management practices.
Originality/value
The value of this review is that it identifies the current trends in tourism and hotel demand modeling and forecasting research and points out future research directions.
This study proposes a relative climate index based on the push and pull theory to assess the effects of relative climate variability on seasonal tourism demand. The relative climate index measures the climatic comfort of a destination relative to that of the tourist origin. Using the proposed approach, the effects of the relative climate comfort on seasonal tourism demand are empirically tested based on a quarterly panel data set of visitor arrivals from Hong Kong to 13 major Chinese cities. The intra-annual seasonality and interannual variability are both tested in the model. The results indicate that the intra-annual relative climate positively influences tourism demand in Mainland regions, where the climate is significantly different from that of Hong Kong.
Hospitality constituencies need accurate forecasting of future performance of hotels in specific destinations to benchmark their properties and better optimize operations. As competition increases, hotel managers have urgent need for accurate short-term forecasts. In this study, time-series models incorporating several tourism big data sources, including search engine queries, website traffic, and weekly weather information, are tested in order to construct an accurate forecasting model of weekly hotel occupancy for a destination. The results show the superiority of ARMAX models with both search engine queries and website traffic data in accurate forecasting. Also, the results suggest that weekly dummies are superior to Fourier terms in capturing the hotel seasonality. The limitations of the inclusion of multiple big data sources are noted since the reduction in forecasting error is minimal.
This study utilizes both disaggregated data and macroeconomic indicators in order to examine the importance of the macroeconomic environment of origin countries for analysing destinations’ tourist arrivals. In particular, it is the first study to present strong empirical evidence that both of these features in tandem provide statistically significant information of tourist arrivals in Greece. The forecasting exercises presented in our analysis show that macroeconomic indicators conducive to better forecasts are mainly origin country-specific, thus highlighting the importance of considering the apparent sharp national contrasts among origin countries when investigating domestic tourist arrivals. Given the extent of the dependency of the Greek economy on tourism income, but also, given the perishable nature of the tourist product itself, results have important implications for policy makers in Greece.
Existing studies on tourism seasonality have been mainly identified at annual or monthly level and the seasonality in tourism demand forecasting has always been addressed by modeling season patterns. However, annual or monthly seasonality is coarse-grained and can’t capture the subtle changes emerging both in tourism theory and practice. This study recognizes tourism seasonality based on intra-day patterns and inter-day similarity and suggests a novel approach to addressing seasonality in tourism demand forecasting. The proposed three-step method contains tourism seasonality recognition, tourism seasonality matching, and tourism demand forecasting. The empirical findings, based on two attractions in China, demonstrate that the proposed method based on dynamic time warping and density-peak clustering can precisely capture tourism seasonality at the daily level. The method can also detect special tourism periods or subtle changes in seasonality, such as staggered peak travel phenomenon. Superior forecasting performance with seasonality matching is also revealed. This study sheds new light on tourism seasonality recognition and contributes to forecasting methodology.
Purpose
This study aims to predict overnight stays in Italy at tourist accommodation facilities through a nonlinear, single factor, stochastic model called CIR#. The contribution of this study is twofold: in terms of forecast accuracy and in terms of parsimony (both from the perspective of the data and the complexity of the modeling), especially when a regular pattern in the time series is disrupted. This study shows that the CIR# not only performs better than the considered baseline models but also has a much lower error than other additional models or approaches reported in the literature.
Design/methodology/approach
Typically, tourism demand tends to follow regular trends, such as low and high seasons on a quarterly/monthly level and weekends and holidays on a daily level. The data set consists of nights spent in Italy at tourist accommodation establishments as collected on a monthly basis by Eurostat before and during the COVID-19 pandemic breaking regular patterns.
Findings
Traditional tourism demand forecasting models may face challenges when massive amounts of search intensity indices are adopted as tourism demand indicators. In addition, given the importance of accurate forecasts, many studies have proposed novel hybrid models or used various combinations of methods. Thus, although there are clear benefits in adopting more complex approaches, the risk is that of dealing with unwieldy models. To demonstrate how this approach can be fruitfully extended to tourism, the accuracy of the CIR# is tested by using standard metrics such as root mean squared errors, mean absolute errors, mean absolute percentage error or average relative mean squared error.
Research limitations/implications
The CIR# model is notably simpler than other models found in literature and does not rely on black box techniques such as those used in neural network (NN) or data science-based models. The carried analysis suggests that the CIR# model outperforms other reference predictions in terms of statistical significance of the error.
Practical implications
The proposed model stands out for being a viable option to the Holt–Winters (HW) model, particularly when dealing with irregular data.
Social implications
The proposed model has demonstrated superiority even when compared to other models in the literature, and it can be especially useful for tourism stakeholders when making decisions in the presence of disruptions in data patterns.
Originality/value
The novelty lies in the fact that the proposed model is a valid alternative to the HW, especially when the data are not regular. In addition, compared to many existing models in the literature, the CIR# model is notably simpler and more transparent, avoiding the “black box” nature of NN and data science-based models.
Since its launch, ChatGPT, an artificial intelligence chatbot developed by Open AI based on the premises of generative pre-trained transformer autoregressive language models, has gained widespread popularity and is making significant impact on society with its unique features, such as natural language processing and contextual awareness. ChatGPT is viewed as a major disruptive innovation that is likely to revolutionize the operations in many industries including the hospitality and tourism industry. The adoption of ChatGPT will result in substantial changes throughout the hospitality and tourism industry by disrupting how customer search for information, make decisions, and how businesses produce, create, and deliver customized services and experiences. This conceptual paper provides a comprehensive discussion on generative pre-trained transformers’ (GPTs) benefits, and potential challenges and threats they pose to the hospitality and tourism industry. The feasibility of integrating GPT into different travel stages and decision-making processes is also discussed. The article concludes by proposing a potential future research agenda on using GPT in creating and delivering hospitality and tourism experiences, which can guide further advancements in the field.
Demand forecasting is important for management and decision making in the tourism sector. However, research on deep learning forecasting, which combines multiple data sources, is still in the development phase. This study proposes a bidirectional long short-term memory (BiLSTM) forecasting method incorporating an attention mechanism (ATT-BiLSTM) that can better extract data features from a set of predictor variables consisting of multiple predictor variables (generated from historical tourist volume, search engine data, weather data and day off data). The research experimentally validates the effectiveness of the method using the famous Chinese tourist attraction Jiuzhaigou as a case study. The results show that the proposed model not only has better generalization ability but also significantly outperforms the four benchmark models, convolutional neural network (CNN), SVR, LSTM, and BiLSTM, in terms of prediction accuracy. In addition, we analyse the importance of the different predictor variables in the prediction model characteristics.
Purpose
This study aims to provide a comprehensive review of hotel demand forecasting to identify its key fundamentals and evolution and future research directions and trends to advance the field.
Design/methodology/approach
Articles on hotel demand modeling and forecasting were identified and rigorously selected using transparent inclusion and exclusion criteria. A final sample of 85 empirical studies was obtained for comprehensive analysis through content analysis.
Findings
Synthesis of the literature highlights that hotel forecasting based on historical demand data dominates the research, and reservation/cancellation data and combined data gradually attracted research attention in recent years. In terms of model evolution, time series and AI-based models are the most popular models for hotel demand forecasting. Review results show that numerous studies focused on hybrid models and AI-based models.
Originality/value
To the best of the authors’ knowledge, this study is the first systematic review of the literature on hotel demand forecasting from the perspective of data source and methodological development and indicates future research directions.
Purpose
Given the importance of spatial effects in improving the accuracy of hotel demand forecasting, this study aims to introduce price and online rating, two critical factors influencing hotel demand, as external variables into the model, and capture the spatial and temporal correlation of hotel demand within the region.
Design/methodology/approach
For high practical implications, the authors conduct the case study in Xiamen, China, where the hotel industry is prosperous. Based on the daily demand data of 118 hotels before and during the COVID-19 period (from January to June 2019 and from January to June 2021), the authors evaluate the prediction performance of the proposed innovative model, that is, a deep learning-based model, incorporating graph convolutional networks (GCN) and gated recurrent units.
Findings
The proposed model simultaneously predicts the daily demand of multiple hotels. It effectively captures the spatial-temporal characteristics of hotel demand. In addition, the features, price and online rating of competing hotels can further improve predictive performance. Meanwhile, the robustness of the model is verified by comparing the forecasting results for different periods (during and before the COVID-19 period).
Practical implications
From a long-term management perspective, long-term observation of market competitors’ rankings and price changes can facilitate timely adjustment of corresponding management measures, especially attention to extremely critical factors affecting forecast demand, such as price. While from a short-term operational perspective, short-term demand forecasting can greatly improve hotel operational efficiency, such as optimizing resource allocation and dynamically adjusting prices. The proposed model not only achieves short-term demand forecasting, but also greatly improves the forecasting accuracy by considering factors related to competitors in the same region.
Originality/value
The originalities of the study are as follows. First, this study represents a pioneering attempt to incorporate demand, price and online rating of other hotels into the forecasting model. Second, integrated deep learning models based on GCN and gated recurrent unit complement existing predictive models using historical data in a methodological sense.
The study analyzes which image repair strategies were adopted by officials and marketers around the world to combat the Covid-19 tourism image crisis. We used qualitative content analysis of news reports, videos/ads, social media posts, and recovery campaigns taken from international and tourism news outlets, websites, and YouTube. As a result, we would like to offer a new theoretical framework that divides the image repair strategies into the three phases of a pandemic tourism crisis: pre-lockdown, lockdown and post-lockdown/recovery. The new theoretical framework might be helpful for studies analyzing destinations’ image repair efforts while combating pandemics, as well for practitioners.
The COVID-19 pandemic has hindered international travel considerably, greatly affecting the hotel industry. Hong Kong, as a well-known international tourist destination, has also been hit hard by the crisis. Recovery forecasts for hotel room demand are critical to managing this ongoing crisis. This study employs the autoregressive distributed lag error correction model to generate baseline forecasts of hotel room demand for Hong Kong followed by compound scenario analysis to optimize forecasts considering the pandemic’s impacts. The COVID-19 Travelable Index is designed to group source markets by their pandemic situations, vaccinations, policy responses, and health resilience. To capture pandemic-related uncertainty, this study presents three scenarios describing recovery patterns based on trough duration, the quarter for lifting travel restrictions, and the quarter for returning to baseline forecasts. Hotel demand forecasts geared toward each source market are analyzed, revealing strategies to help hotel businesses manage this crisis.
Forecasting tourist demand for multiple tourist attractions on an hourly basis provides important insights for tourism management effectiveness and efficiency, such as staffing and resource optimization. However, hourly data is dynamic and nonlinear and cannot be handled by using existing forecasting models. This study develops an improved artificial intelligent-based model, Correlated Time Series oriented Long Short-Term Memory with Attention Mechanism, to solve this problem. The model compensates for the lack of spatial dependence characteristics between sequences in previous models. The validity of the model is verified through a forecasting exercise for 77 attractions in Beijing, China. The results show that our model significantly outperforms the baseline models. Our model can help managers improve resource utilization efficiency and staff and tourist satisfaction.
In a context in which the tourism industry is jeopardised by the COVID-19 pandemic, and potentially by other pandemics in the future, the capacity to produce accurate forecasts is crucial to stakeholders and policy-makers. This paper attempts to forecast the recovery of tourism demand for 2021 in 20 destinations worldwide. An original scenario-based judgemental forecast based on the definition of a Covid-19 Risk Exposure index is proposed to overcome the limitations of traditional forecasting methods. Three scenarios are proposed, and ex ante forecasts are generated for each destination using a baseline forecast, the developed index and a judgemental approach. The limitations and potential developments of this new forecasting model are then discussed.
This paper is to produce different scenarios in forecasts for international tourism demand, in light of the COVID-19 pandemic. By implementing two distinct methodologies (the Long Short Term Memory neural network and the Generalized Additive Model), based on recent crises, we are able to calculate the expected drop in the international tourist arrivals for the next 12 months. We use a rolling-window testing strategy to calculate accuracy metrics and show that even though all models have comparable accuracy, the forecasts produced vary significantly according to the training data set, a finding that should be alarming to researchers. Our results indicate that the drop in tourist arrivals can range between 30.8% and 76.3% and will persist at least until June 2021.
Internet techniques significantly influence the tourism industry and Internet data have been used widely used in tourism and hospitality research. However, reviews on the recent development of Internet data in tourism forecasting remain limited. This work reviews articles on tourism forecasting research with Internet data published in academic journals from 2012 to 2019. Then, the findings ae synthesized based on the following Internet data classifications: search engine, web traffic, social media, and multiple sources. Results show that among such classifications, search engine data are most widely incorporated into tourism forecasting. Time series and econometric forecasting models remain dominant, whereas artificial intelligence methods are still developing. For unstructured social media and multi-source data, methodological advancements in text mining, sentiment analysis, and social network analysis are required to transform data into time series for forecasting. Combined Internet data and forecasting models will help in improving forecasting accuracy further in future research.
Traditional tourism demand forecasting models may face challenges when massive amounts of search intensity indices are adopted as tourism demand indicators. Using a deep learning approach, this research studied the framework in forecasting monthly Macau tourist arrival volumes. The empirical results demonstrated that the deep learning approach significantly outperforms support vector regression and artificial neural network models. Moreover, the construction and identification of highly relevant features from the proposed deep network architecture provide practitioners with a means of understanding the relationships between various tourist demand forecasting factors and tourist arrival volumes.
This article also launches the Annals of Tourism Research Curated Collection on Tourism Demand Forecasting, a special selection of research in this field
Highlights
1.Our AI-based deep-learning approach contributes to higher forecasting accuracy.
2. Our innovative deep-learning model alleviates the limited data availability.
3. With our new pooling method, we significantly reduce model overfitting.
4. We reveal similar cross-country demand patterns for the Asia-Pacific regions.
Abstract
Advances in tourism demand forecasting immensely benefit tourism and other sectors, such as economic and resource management studies. However, even for novel AI-based methodologies, the challenge of limited available data causing model overfitting and high complexity in forecasting models remains a major problem. This study proposes a novel group-pooling-based deep-learning model (GP–DLM) to address these problems and improve model accuracy. Specifically, with our group-pooling method, we advance the tourism forecasting literature with the following findings. First, GP–DLM provides superior accuracy in comparison with benchmark models. Second, we define the novel dynamic time warping (DTW) clustering quantitative approach. Third, we reveal cross-region factors that influence travel demands of the studied regions, including “travel blog,” “best food,” and “Air Asia.”
This study reviewed 72 studies in tourism demand forecasting during the period from 2008 to 2017. Forecasting models are reviewed in three categories: econometric, time series and artificial intelligence (AI) models. Econometric and time series models that have already been widely used before 2007 remained their popularity and were more often used as benchmark models for forecasting performance evaluation and comparison with respect to new models. AI models are rapidly developed in the past decade and hybrid AI models are becoming a new trend. And some new trends with regard to the three categories of models have been identified, including mixed frequency, spatial regression and combination and hybrid models. Different combination components and combination techniques have been discussed. Results in different studies proved superiority of combination forecasts over average single forecasts performance.
This study aims to evaluate the accuracy of different judgmental forecasting tasks, compare the judgmental forecasting behaviour of tourism researchers and practitioners and explore the validity of experts’ judgmental behaviour by using the Hong Kong visitor arrivals forecasts over the period 2011Q2−2015Q4. Delphi-based judgmental forecasting procedure was employed through the Hong Kong Tourism Demand Forecasting System, an online forecasting support system, to collect and combine experts’ adjusted forecasts. This study evaluates forecasting performance and explores the characteristics of judgmental adjustment behaviour through the use of a group of error measures and statistical tests. The findings suggest a positive correlation between forecast accuracy and the level of data variability, and that experts’ adjustments are more beneficial in terms of achieving higher accuracy for series with higher variability. Industry practitioners’ forecasts outperformed academic researchers, particularly in making short-term forecasts. However, no significant difference was found between the two panels in making directionally correct forecasts. Experts’ judgmental intervention was found most useful for those series most in need of adjustment. The size of adjustment was found to have a strong and significantly positive association with the direction of forecast adjustment, but no statistically significant evidence was found regarding the relationship between accuracy improvement and adjustment size.
