Multi-Horizon Forecasting of Monthly Air Cargo Throughput at Incheon International Airport Using the Temporal Fusion Transformer
인천국제공항 월별 항공 화물 처리량의 다중 시계열 예측 : Temporal Fusion Transformer를 활용한 분석
This study evaluates the Temporal Fusion Transformer (TFT) for forecasting monthly international air cargo throughput at Incheon International Airport. Since air cargo demand is influenced by temporal dependence, seasonality, macroeconomic conditions, cost factors, and external disruptions, the problem is formulated as a multivariate multi-horizon forecasting task. A TFT-based framework is implemented using monthly data, incorporating past target sequences, observed past inputs, and known future inputs to preserve temporal realism, and is compared with SARIMA, vanilla LSTM, and Transformer using MAE, RMSE, and MAPE. The results show that TFT achieves the highest forecasting accuracy and exhibits a more stable error profile over the 12-month horizon, while providing informative prediction intervals and interpretable results, with lagged cargo throughput, global trade conditions, and industrial production identified as key predictors, supporting its practical usefulness for operational planning.
한국어
본 연구는 인천국제공항의 월별 국제 항공 화물 처리량 예측을 위해 Temporal Fusion Transformer(TFT) 의 성능을 평가한다. 항공 화물 수요는 시간적 의존성, 계절성, 거시경제 여건, 비용 요인, 그리고 외생적 충격의 영향을 받기 때문에, 본 문제는 다변량 다중 시계열 예측 문제로 설정된다. 이를 위해 월별 데이터를 기반으로 TFT 기반 예측 프레임워크를 구축하였으며, 과거 목표 변수, 관측된 과거 입력 변수, 그리고 알려진 미래 입력 변수를 포함하여 시간적 타당성을 유지하였다. 또한 SARIMA, vanilla LSTM, Transformer 모형과 비교하여 MAE, RMSE, MAPE 기준으로 성능을 평가하였다. 분석 결과, TFT는 모든 비교 모형 중 가장 높은 예측 정확 도를 보였으며, 12개월 예측 구간 전반에 걸쳐 보다 안정적인 오차 패턴을 나타냈다. 또한 예측 구간과 해석 가능 한 결과를 제공하였으며, 시차 화물 처리량, 글로벌 교역 여건, 산업 생산 지수가 주요 영향 변수로 확인되었다. 이러한 결과는 TFT가 공항 운영 계획을 위한 실질적으로 유용한 예측 프레임워크임을 시사한다.
목차
Abstract 요약 Ⅰ. Introduction Ⅱ. Theoretical Background 2.1 Methodological Evolution in Air Cargo Demand Forecasting 2.2 Overview of the Temporal Fusion Transformer Ⅲ. Methodology 3.1 Temporal Fusion Transformer-based Forecasting Framework 3.2 Data Description and Preprocessing 3.3 Forecasting Configuration and Training Strategy Ⅳ. Forecasting Results and Analysis 4.1 Comparative Evaluation of Overall Forecast Accuracy 4.2 Temporal Forecast Accuracy over the Forecast Horizon 4.3 Forecast Visualization and Uncertainty Analysis 4.4 Interpretability Analysis 4.4 Interpretability Analysis Ⅴ. Conclusion 5.1 Summary of Contributions 5.2 Limitations and Future Directions REFERENCES
키워드
항공 화물 수요 예측인천국제공항Temporal Fusion Transformer다중 시계열 예측해석 가능성Air Cargo Demand ForecastingIncheon International AirportTemporal Fusion TransformerMulti-horizon ForecastingInterpretability
저자
Kyeongmin Yum [ 염경민 | Assistant Professor, Department of Global Business, Tongmyong University ]
Corresponding Author
Ever since next generation convergence technology became one of the most important industries in the nation, computing professionals have encountered a growing number of challenges. Along with scholars and colleagues in related fields, they have gathered in avariety of forums and meetings over the last few decades to share their knowledge, experiences and the outcome of their research. These exchanges have led to the founding of the International Next-generation Convergence technology (INCA) on December 1, 2015. INCA was registered as an incorporated association under the Ministry of Information and Communications. The main purpose of the organization is to improve our society by achieving the highest capability possible in next generation convergence technology.
간행물
간행물명
차세대융합기술학회논문지 [The Journal of Next-generation Convergence Technology Association]