Air passenger forecasting using Neural Granger causal Google trend queries

doi:10.1016/j.jairtraman.2021.102083

Journal of Air Transport Management

Volume 95, August 2021, 102083

https://doi.org/10.1016/j.jairtraman.2021.102083 Get rights and content

Highlights

•
A novel method based on Google Trend Queries is proposed to identify internet search queries that can forecast air passengers.
•
171 Neural Granger Causal Google Trends Search queries are identified out of an initial 1317 queries using a word2vec model.
•
Neural Granger Queries inputs to a forecasting model produced a higher forecasting performance.

Abstract

Air passenger forecasting provides important insights for both Governments and Aerospace industries to plan their for their future activities. Google Trends can provide a large database of historical search query frequency which can be used as explanatory variables for air passenger forecasting. This paper explores the use of a Neural Granger Causality model to select the best search query that can forecast arrival air passengers in Singapore Changi Airport. Neural Granger Causality models are an extension of the original Granger Causality model that uses neural networks instead of Linear Vector Auto-Regressive (VAR) models to capture non-linear relations between the targets and the tested explanatory variables. In this paper, 1317 Google Trends search queries are tested for Neural Granger Causality of which 171 queries are deemed as Neural Granger Causal for forecasting Singapore Changi Airport monthly arrival passengers. The model that used all 171 Neural Granger Queries achieved the highest $R^{2}$ value ( $R^{2} = 0.919$ ) with the lowest Standard Deviation ( $S D = 0.363$ ) compared to the other models which was not filtered for Neural Granger Causality. The 171 queries found are search terms that reflects a unidirectional neural granger causal relationship with the number of arrival air passengers at Changi Airport.

Introduction

The air transportation industry has witnessed tremendous growth over the past decades. Significant infrastructural developments have hence followed to support this growth and accommodate the ever increasing passenger numbers. The International Air Transport Association (IATA) forecast predicts 8.2 billion passengers worldwide by the year 2037, at an annual compounded growth rate of 3.5% IATA (2019). This long term passenger growth forecast by IATA stayed the same after a revision due to the Covid-19 pandemic IATA (2020). Consequently, it has become imperative to develop tools and methodologies to efficiently generate short and long term passenger traffic forecasts at airports. Such forecasts aim to assist airside and landside operational planning, short term maintenance plans and flight schedules. Over the long term, these forecasts also assist in the future planning of Airport infrastructure, assist airline companies in equipment purchase and route structure and assist aircraft manufacturers to design future aircraft that are optimally profitable to their customers (ICAO (2006) and Kim et al. (2016)).

Traditionally, econometric predictors (GDP, Oil Price etc) are used for air passenger forecasts but calculation of the econometric predictors is generally tedious. Alternatively, data from internet search queries provides a source of predictors that are up to date, requires little post-processing and reflects the general population psychology of users (Bragazzi (2014)). However, the large number of search terms would mean that the best search terms will need to be identified if search query data were to be used for air passenger forecasting. If the best search queries are identified, they could also be used for further causality analysis which may be useful for policy makers in the aviation industry.

The purpose of this research is to: 1) Find the relevant search queries that are the most useful predictors for air passenger forecasting at Singapore Changi Airport using Neural Granger Causality (NGC) analysis which is adapted from Tank et al. (2018). 2) Investigate the efficacy of using neural granger causal search query trends as predictors to forecast the number of air passengers.

This paper will be structured as follows: Section 2 will highlight several related works and its relation to the research in this paper. Section 3 will describe the methodologies used to collect data and the describe the details of Neural Granger Causality. The results will then be presented in Section 6 a feasibility study of NGC analysis will be demonstrated on the synthetic Lorenz-96 dataset and the Changi/Google Trends dataset. Section 7 will discuss on the applicability of NGC analysis in the post-covid era and finally, this paper will conclude in Section 8.

Section snippets

Literature review

Econometric variables has been used as predictors for air passenger forecast before. Fernandes and Pacheco (2010) used a linear granger causality test to determine causality between GDP and domestic air passenger traffic in Brazil and concluded that there is an uni-directional causal relationship between econometric growth to domestic air transport demand. Hakim and Merkert (2016) also examined the causal relationship between air transport and econometric growth in the south-east asian region

Finding initial queries

One of the objectives of this research is to find the relevant search query predictors to forecast the number of Changi Airport air passengers. Even if the search queries are constrained to a single language (In this case search terms in English), there will be a large amount of possible search queries and enumerating through all search queries will not be an efficacious solution to find the inital search queries for Neural Granger Causality analysis. A smarter solution would be to use a

Analysis of target dataset - main characteristics of air passenger numbers

The target outputs (Number of Air Passengers) of the forecasting model in this study are obtained from Govtech (2020). Each numerical entry in the dataset represents the total number of air passenger arrivals in 1 month and this data set is in the range from January 1961 to Jan 2020. A seasonal and trend decomposition of this data are illustrated in Fig. 3. The seasonal decomposition reveals that the air passenger data shows that there is a repeating cycle every 4 Months. The residuals plot

Lorenz-96

The Lorenz-96 model is commonly used for data assimilation Lorenz (1996) and in this case can be used to generate time series data. Since the target variables can be artificially constructed, we can test the efficacy of Neural Granger Causal analysis on this synthetic dataset (by integrating Equation (7)) to find out if it can identify the original predictors. An experiment was set up to with 10 variables ( $x_{1}, x_{2}, x_{3} \dots . x_{10}$ ) and the forcing value F was set to 5. η was set to 0.8 and λ was set to

Proof of concept on Lorenz-96 synthetic dataset

The plots in Fig. 5 and Fig. 6 highly suggests that the neural granger model can correctly identify potentially causal time series as the weight values of the input variables ( $x_{4}, x_{5}, x_{6}$ ) has a significantly higher value than the other variables. Weights sparsity can also be observed when comparing the weights of the input time lags (Equation (8)) shown in Fig. 7 suggesting that the neural granger causality analysis can identify the best features and its corresponding time lag.

Both $L_{1}$ and SGL

Discussion - applicability in air traffic management

The results shown by comparing Table 4, Table 5 suggests that the search queries which are filtered by Neural Granger Causality analysis are specifically important search terms that may have a causal relationship with the number of air passengers in Changi Airport. One interesting characteristic of the neural granger queries (Table 3) is that other south-east asian vacation/airports shows up as Neural Granger Causal (“Redang Island”, “Villas Hua Hin”, “Kuta Bali”, “Soekarno Hatta Airport”,

Conclusion

The purpose of this paper is to investigate and evaluate the effects of using neural granger causal Google Trends queries to forecast arrival passengers in Changi International Airport, Singapore. Neural Granger Causality models employ the use of component-wise neural networks with sparse inducing regularization parameters to identify the best predictors for forecasting a time-series.

The experimental results strongly suggests that using the Neural Granger Causality model can identify Google

References (23)

D. Baker et al.
Regional aviation and economic growth: cointegration and causality analysis in Australia
J. Transport Geogr.
(2015)
T. Grosche et al.
Gravity models for airline passenger volume estimation
J. Air Transport. Manag.
(2007)
M.M. Hakim et al.
The causal relationship between air transport and economic growth: empirical evidence from south asia
J. Transport Geogr.
(2016)
S. Kim
Forecasting short-term air passenger demand using big data from search engine queries
Autom. ConStruct.
(2016)
S. Sun et al.
Nonlinear vector auto-regression neural network for forecasting air passenger flow
J. Air Transport. Manag.
(2019)
E. Suryani et al.
Air passenger demand forecasting and passenger terminal capacity expansion: a system dynamics framework
Expert Syst. Appl.
(2010)
N.L. Bragazzi
A google trends-based approach for monitoring nssi
Psychol. Res. Behav. Manag.
(2014)
S.-C. Chen et al.
Improving the forecasting accuracy of air passenger and air cargo demand: the application of back-propagation neural networks
Transport. Plann. Technol.
(2012)
H. Choi et al.
Predicting the present with google trends
Econ. Rec.
(2012)
E. Fernandes et al.
The causal relationship between gdp and domestic air passenger traffic in Brazil
Transport. Plann. Technol.
(2010)

Google

Google Code Archive - Long-term Storage for Google Code Project Hosting

(2013)

Cited by (16)

Comparison of artificial neural networks and regression analysis for airway passenger estimation
2024, Journal of Air Transport Management
With the increasing demand in operations, time is getting more important. In order to use time and energy more effectively, it is becoming more important for airline companies and airport managements to make strategic plans for the future. To make beneficial and correct strategic plans for airways, one of the factors that is needed to be considered is future passenger numbers. With more accurate passenger number forecasts, airport managements can act more efficiently and reduce time, energy consumption and hence would be able to reduce costs. In this study, airway passenger number estimation is handled. Three metropolitan cities’ airport passenger numbers are considered. Artificial neural networks and regression analysis are carried out to estimate passenger number. In addition, data are handled in two different ways. Firstly, ANN and regression analysis are applied using original data series. In the second step, seasonal decomposition is applied on the data series and both approaches are repeated for deseasonal series. In Artificial Neural Networks approach, an experimental design is developed considering training algorithms, number of input nodes and number of nodes in the hidden layer which make up 960 design points. In the results of these experiments, performance of ANN approach is tested for three input factors and high-performance design points are identified. Furthermore, for benchmarking purposes, regression analysis is carried out. Linear, logarithmic, power, exponential, and polynomial models are developed. Finally, results of ANN and regression approaches are compared in terms of mean absolute percent error, and it is found that ANN overperformed compared to regression analysis.
Seasonal variations in daily data: An application to air passenger arrivals
2023, Journal of Air Transport Management
The aim of this paper is to describe and compare seasonal effects in daily air passenger arrivals. Multiple seasonal cycles of different lengths are usually observed in daily time series. To model weekly cycles over a short and fixed seasonal period, conventional formulations can be applied. However, the length of yearly and, above all, monthly seasonal cycles is not fixed. Yet, whatever the lengths of the seasonal periods, seasonal cycles can be modelled by using evolving spline functions in such a way that a seasonal effect at a proportion of the seasonal period is defined as a non-fixed parametric formulation of this proportion. Thus, restricted evolving spline models are useful to model this type of seasonal variation. The areas under the splines can measure changes in the magnitude of seasonal variations over time and can be used to compare the relevance of seasonal variations for different seasonal periods or different time series. This proposal is illustrated by applying it to daily series of air passengers arriving in the Canary Islands and at the rest of Spain's airports.
Airport terminal passenger forecast under the impact of COVID-19 outbreaks: A case study from China
2023, Journal of Building Engineering
Citation Excerpt :
Additional socioeconomic factors, such as national and regional gross domestic product (GDP), income levels, population, employment rates, and special events, were used to fine-tune the forecast [21,25]. Some studies have also attempted to predict travel demands with unconventional factors such as search engine queries [26,27] and social media data [28]. However, the outbreak of the COVID-19 pandemic has substantially increased the uncertainty of air travel demand, an effect that is difficult to address with the traditional predictors described above.
Passengers significantly affect airport terminal energy consumption and indoor environmental quality. Accurate passenger forecasting provides important insights for airport terminals to optimize their operation and management. However, the COVID-19 pandemic has greatly increased the uncertainty in airport passenger since 2020. There are insufficient studies to investigate which pandemic-related variables should be considered in forecasting airport passenger trends under the impact of COVID-19 outbreaks. In this study, the interrelationship between COVID-19 pandemic trends and passenger traffic at a major airport terminal in China was analyzed on a day-by-day basis. During COVID-19 outbreaks, three stages of passenger change were identified and characterized, i.e., the decline stage, the stabilization stage, and the recovery stage. A typical “sudden drop and slow recovery” pattern of passenger traffic was identified. A LightGBM model including pandemic variables was developed to forecast short-term daily passenger traffic at the airport terminal. The SHapley Additive exPlanations (SHAP) values was used to quantify the contribution of input pandemic variables. Results indicated the inclusion of pandemic variables reduced the model error by 27.7% compared to a baseline model. The cumulative numbers of COVID-19 cases in previous weeks were found to be stronger predictors of future passenger traffic than daily COVID-19 cases in the most recent week. In addition, the impact of pandemic control policies and passengers' travel behavior was discussed. Our empirical findings provide important implications for airport terminal operations in response to the on-going COVID-19 pandemic.
Explainable AI for Operational Research: A defining framework, methods, applications, and a research agenda
2023, European Journal of Operational Research
The ability to understand and explain the outcomes of data analysis methods, with regard to aiding decision-making, has become a critical requirement for many applications. For example, in operational research domains, data analytics have long been promoted as a way to enhance decision-making. This study proposes a comprehensive, normative framework to define explainable artificial intelligence (XAI) for operational research (XAIOR) as a reconciliation of three subdimensions that constitute its requirements: performance, attributable, and responsible analytics. In turn, this article offers in-depth overviews of how XAIOR can be deployed through various methods with respect to distinct domains and applications. Finally, an agenda for future XAIOR research is defined.
Detecting the hidden asymmetric relationship between crude oil and the US dollar: A novel neural Granger causality method
2023, Research in International Business and Finance
The relationship between financial series is not always easy to detect due to their underlying asymmetry and nonlinearity. Both characteristics are not usually considered simultaneously, which may lead to many drawbacks in financial analysis. Hence, we develop a novel neural Granger causality method from both asymmetric and nonlinear perspectives and further revisit the response and impact of crude oil on the exchange rate. Our findings reveal the unidirectional nonlinear and asymmetric effect of crude oil on the exchange rate; that is, positive and negative oil prices can have a substantial impact on exchange rate shocks. Interestingly, this influence seems to strengthen after the Russia–Ukraine conflict. Besides, we also use simulation technology to evaluate the rationality and effectiveness of our proposed methods. Investors, policymakers, and scholars may be interested in our findings regarding the oil-dollar relationships; as well as interested in applying our methodology to other contexts.
Copula ARMA-GARCH modelling of spatially and temporally correlated time series data for transportation planning use
2023, Transportation Research Part C: Emerging Technologies
Time series analysis has been used extensively in transport research in various areas, such as traffic management and transport planning. Time-series data may contain temporal and spatial correlations. Temporal correlation refers to the dependency of observations on time, while spatial correlation refers to the dependence of observations on space. Although there is a substantial amount of literature on statistical models that modelled temporal correlation of transport time-series data, there are few studies that accommodate both temporal and spatial correlations. In cases where more than one time-series dataset is collected from different locations, each observation can depend on its previous values and observations from other locations. Transport-related time-series data (e.g., traffic volumes, travel times, etc.) may exhibit nonlinear spatial correlation within the transport network. However, existing statistical time series models that consider spatial correlation, such as vector auto-regression (VAR) models, are limited to linear spatial correlation functions with a time-invariance assumption. This study introduces a copula time series model capable of modelling correlations between variables through time and space. Temporal correlations are modelled by an autoregressive moving average-generalised autoregressive conditional heteroskedasticity (ARMA-GARCH) model, where the conditional mean is predicted based on past values and errors, and conditional variance is predicted based on past residuals and conditional variances. This enables the model to describe heteroscedastic data. The spatial correlation is modelled with a copula model, which has the flexibility to model different types of correlations, such as nonlinear, tailed and asymmetric correlations. We illustrate the statistical properties of our approach where the performance of the copula ARMA-GARCH model is compared with the VAR model based on mean absolute percentage error (MAPE) and root mean square error (RMSE) values of model predictions on synthesised data. We then demonstrate practical implementations of these models examined in two case studies with traffic count data collected from four main arterial roads in Sydney.

View all citing articles on Scopus

View full text

Air passenger forecasting using Neural Granger causal Google trend queries

Highlights

Abstract

Introduction

Section snippets

Literature review

Finding initial queries

Analysis of target dataset - main characteristics of air passenger numbers

Lorenz-96

Proof of concept on Lorenz-96 synthetic dataset

Discussion - applicability in air traffic management

Conclusion

J. Transport Geogr.

J. Air Transport. Manag.

J. Transport Geogr.

Autom. ConStruct.

J. Air Transport. Manag.

Expert Syst. Appl.

A google trends-based approach for monitoring nssi

Psychol. Res. Behav. Manag.

Improving the forecasting accuracy of air passenger and air cargo demand: the application of back-propagation neural networks

Transport. Plann. Technol.

Predicting the present with google trends

Econ. Rec.

The causal relationship between gdp and domestic air passenger traffic in Brazil

Transport. Plann. Technol.

Google Code Archive - Long-term Storage for Google Code Project Hosting