Accurate air cargo volume (ACV) forecasting is critical for operational management and global supply chain decision optimization in aviation logistics. Existing methods present limitations in capturing the multi-period seasonal dependencies of ACV sequences and high-order nonlinear interactions between multi-source influencing factors, resulting in insufficient forecasting accuracy under small-sample monthly data scenarios. To address these gaps, this study proposes ACVPNet, a lightweight multi-layer perceptron (MLP)-based ACV forecasting framework integrating multi-period time dependency modeling and multivariate feature interaction learning. The proposed model adopts a dual-branch prediction architecture that independently models cyclical and trend components. By incorporating specialized modules to jointly capture intra-period and inter-period dependencies, it effectively integrates cross-feature information, thereby enhancing both the flexibility and interpretability of temporal representation learning. Specifically, ACVPNet integrates two core modules: the temporal local multi-layer perceptron (TLM), which extracts nonlinear temporal dependencies within and across cycles through localized patch-based learning, and the feature interaction MLP (FIM), which captures inter-variable correlations among economic, consumption, and energy indicators. Together, these components enable ACVPNet to learn complex periodic patterns and multi-dimensional relationships from heterogeneous time series data. Experiments based on the CEIC historical China ACV dataset demonstrate the superiority of the proposed model over three representative baseline models: LSTM, Informer, and SARIMA. In the one-year-ahead ACV forecasting task, ACVPNet achieved an RMSE of 0.083 million tons, representing reductions of 37.6%, 49.7%, and 39.4% compared with the three baseline models, thereby demonstrating a significant performance advantage. The results confirm that ACVPNet provides high-precision and temporally consistent ACV forecasts, offering valuable insights for aviation market analysis, capacity planning, and policy formulation.
Citation: Xiaoyuan Cheng. ACVPNet: An air cargo volume prediction model for China based on periodic structure and multi-dimensional feature interaction[J]. Journal of Industrial and Management Optimization, 2026, 22(5): 2428-2450. doi: 10.3934/jimo.2026089
Accurate air cargo volume (ACV) forecasting is critical for operational management and global supply chain decision optimization in aviation logistics. Existing methods present limitations in capturing the multi-period seasonal dependencies of ACV sequences and high-order nonlinear interactions between multi-source influencing factors, resulting in insufficient forecasting accuracy under small-sample monthly data scenarios. To address these gaps, this study proposes ACVPNet, a lightweight multi-layer perceptron (MLP)-based ACV forecasting framework integrating multi-period time dependency modeling and multivariate feature interaction learning. The proposed model adopts a dual-branch prediction architecture that independently models cyclical and trend components. By incorporating specialized modules to jointly capture intra-period and inter-period dependencies, it effectively integrates cross-feature information, thereby enhancing both the flexibility and interpretability of temporal representation learning. Specifically, ACVPNet integrates two core modules: the temporal local multi-layer perceptron (TLM), which extracts nonlinear temporal dependencies within and across cycles through localized patch-based learning, and the feature interaction MLP (FIM), which captures inter-variable correlations among economic, consumption, and energy indicators. Together, these components enable ACVPNet to learn complex periodic patterns and multi-dimensional relationships from heterogeneous time series data. Experiments based on the CEIC historical China ACV dataset demonstrate the superiority of the proposed model over three representative baseline models: LSTM, Informer, and SARIMA. In the one-year-ahead ACV forecasting task, ACVPNet achieved an RMSE of 0.083 million tons, representing reductions of 37.6%, 49.7%, and 39.4% compared with the three baseline models, thereby demonstrating a significant performance advantage. The results confirm that ACVPNet provides high-precision and temporally consistent ACV forecasts, offering valuable insights for aviation market analysis, capacity planning, and policy formulation.
| [1] |
J. G. M. Anguita, O. D. Olariaga, Air cargo transport demand forecasting using ConvLSTM2D, an artificial neural network architecture approach, Case Stud. Transp. Policy, 12 (2023), 101009. https://doi.org/10.1016/j.cstp.2023.101009 doi: 10.1016/j.cstp.2023.101009
|
| [2] |
C. C. Hwang, G. C. Shiao, Analyzing air cargo flows of international routes: an empirical study of Taiwan Taoyuan International Airport, J. Transp. Geogr., 19 (2011), 738-744. https://doi.org/10.1016/j.jtrangeo.2010.09.001 doi: 10.1016/j.jtrangeo.2010.09.001
|
| [3] |
F. Kupfer, H. Meersman, E. Onghena, E. V. Voorde, The underlying drivers and future development of air cargo, J. Air Transp. Manag., 61 (2017), 6-14. https://doi.org/10.1016/j.jairtraman.2016.07.002 doi: 10.1016/j.jairtraman.2016.07.002
|
| [4] |
W. W. L. Lo, Y. Wan, A. Zhang, Empirical estimation of price and income elasticities of air cargo demand: The case of Hong Kong, Transp. res., Part A Policy pract., 78 (2015), 309-324. https://doi.org/10.1016/j.tra.2015.05.014 doi: 10.1016/j.tra.2015.05.014
|
| [5] |
Y. Rodríguez, O. D. Olariaga, Air traffic demand forecasting with a bayesian structural time series approach, Period. Polytech. Transp. Eng., 52 (2024), 75-85. https://doi.org/10.3311/PPtr.20973 doi: 10.3311/PPtr.20973
|
| [6] |
D. C. Han, Y. Y. Peng, Prediction of air freight volume based on BP neural network, Int. Conf. Electron. Inf. Technol. Comput. Eng., 4 (2024), 904-907. https://doi.org/10.1145/3650400.3650553 doi: 10.1145/3650400.3650553
|
| [7] |
Q. H. Nguyen, Modeling the volatility of international air freight: A case study of Singapore using the SARIMAX-EGARCH model, J. Air Transp. Manag., 117 (2024), 102593. https://doi.org/10.1016/j.jairtraman.2024.102593 doi: 10.1016/j.jairtraman.2024.102593
|
| [8] |
J. M. Liu, L. N. Ding, X. Y. Guan, J. Gui, J. B. Xu, Comparative analysis of forecasting for air cargo volume: Statistical techniques vs. machine learning, J. Data Inf. Manag., 2 (2020), 243-255. https://doi.org/10.1007/s42488-020-00031-1 doi: 10.1007/s42488-020-00031-1
|
| [9] |
Q. H. Nguyen, P. Q. Tran, P. D. Ngo, Air cargo traffic forecasting model: An empirical study in Vietnam using the SARIMA-X/(E)GARCH model, Res. Transp. Bus. Manag., 59, (2025), 101268. https://doi.org/10.1016/j.rtbm.2024.101268 doi: 10.1016/j.rtbm.2024.101268
|
| [10] |
K. C. Min, H. K. Ha, Forecasting the daily demand of air cargo using data mining with CHAID approach, J. Korean Soc. Transp., 38 (2020), 190-207. https://doi.org/10.7470/jkst.2020.38.3.190 doi: 10.7470/jkst.2020.38.3.190
|
| [11] |
H. Shin, G. Lee, Factors Affecting Air Cargo Demand: Focus on Trade Volumes between South Korea and Intra-Asia, J. Korean Acad. Int. Commer., 34 (2019), 191-214. https://doi.org/10.18104/kaic.2019.34.3.191 doi: 10.18104/kaic.2019.34.3.191
|
| [12] |
H. T. Li, J. C. Bai, X. Cui, Y. W. Li, S. L. Sun, A new secondary decomposition-ensemble approach with cuckoo search optimization for air cargo forecasting, Appl. Soft Comput., 90 (2020), 106161. https://doi.org/10.1016/j.asoc.2020.106161 doi: 10.1016/j.asoc.2020.106161
|
| [13] |
B. Z. Niu, Z. P. Dai, X. P. Zhuo, Co-opetition effect of promised-delivery-time sensitive demand on air cargo carriers' big data investment and demand signal sharing decisions, Transp. Res. E: Logist. Transp. Rev., 123 (2019), 29-44. https://doi.org/10.1016/j.tre.2019.01.011 doi: 10.1016/j.tre.2019.01.011
|
| [14] |
X. Q. Hu, R. M. Jia, Y. Q. Wang, Research on Chengdu air cargo forecast based on improved ARIMA-GARCH, Int. J. Model. Oper. Manag., 8 (2021), 299-312. https://doi.org/10.1504/IJMOM.2021.116802 doi: 10.1504/IJMOM.2021.116802
|
| [15] | T. Y. Chou, G. S. Liang, T. C. Han, Application of fuzzy regression on air cargo volume forecast, Qual. Quant., 47, (2013), 897-908. https://doi.org/10.1007/s11135-011-9572-4 |
| [16] | C. Çatuk, Forecasting Turkey's air cargo tonnage: A comparative analysis of statistical techniques and machine learning methods, J. Article, 9 (2025), 109-117. https://doi.org/10.30518/jav.1582814 |
| [17] |
J. X. Che, W. X. Xia, Y. F. Xu, K. Hu, Multivariate wind speed forecasting with genetic algorithm-based feature selection and oppositional learning sparrow search, Inf. Sci., 695 (2025), 121736. https://doi.org/10.1016/j.ins.2024.121736 doi: 10.1016/j.ins.2024.121736
|
| [18] |
M. S. Alam, J. B. Deb, A. A. Amin, S. Chowdhury, An artificial neural network for predicting air traffic demand based on socio-economic parameters, Decis. Anal. J., 10 (2024), 100382. https://doi.org/10.1016/j.dajour.2023.100382 doi: 10.1016/j.dajour.2023.100382
|
| [19] |
T. Y. Du, Q. W. Yin, Study on freight volume prediction of routes based on random forest model, TCSISR, 5 (2024), 1733-1739. https://doi.org/10.62051/fedtxr69 doi: 10.62051/fedtxr69
|
| [20] | S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural Comput., 9 (1997), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735 |
| [21] |
A. Neagoe, E. I. Tică, L. I. Vuță, O. Nedelcu, G. E. Dumitran, B. Popa, Hybrid LSTM-ARIMA model for improving multi-step inflow forecasting in a reservoir, Water, 17 (2025), 3051. https://doi.org/10.3390/w17213051 doi: 10.3390/w17213051
|
| [22] | Y. W. Wei, W. Y. Tao, H. R. Zhang, D. B. Kou, Research on cargo volume forecasting based on the ARIMA-LSTM hybrid model, ICDACAI, 4 (2024) 896-900. https://doi.org/10.1109/ICDACAI65086.2024.00169 |
| [23] |
S. Q. Luo, T. H. Jiao, X. Y. Zhang, Y. T. Liu, R. R. Li, M. H. Guo, Logistics cargo volume prediction model based on combined ARIMA-LSTM prediction methods, IACIS, 4 (2024), 1-8. https://doi.org/10.1109/IACIS61494.2024.10721663 doi: 10.1109/IACIS61494.2024.10721663
|
| [24] |
F. Kupfer, H. Meersman, E. Onghena, E. V. Voorde, The underlying drivers and future development of air cargo, J. Air Transp. Manag., 61 (2017), 6-14. https://doi.org/10.1016/j.jairtraman.2016.07.002 doi: 10.1016/j.jairtraman.2016.07.002
|
| [25] |
M. S. Devi, F. Garestier, S. Hosseini, Legendre polynomials for nonlinear modeling in InSAR time series, IEEE Geosci. Remote Sens. Lett., 22 (2025), 1-5. https://doi.org/10.1109/LGRS.2025.3570030 doi: 10.1109/LGRS.2025.3570030
|
| [26] |
G. P. Zeng, A unified definition of mutual information with applications in machine learning, Math. Probl. Eng., 2 (2015), 201874. https://doi.org/10.1155/2015/201874 doi: 10.1155/2015/201874
|
| [27] | Z. Li, S. Y. Qi, Y. D. Li, Z. L. Xu, Revisiting long-term time series forecasting: An investigation on linear mapping, arXiv preprint arXiv: 2305.10721, 2023. https://doi.org/10.48550/arXiv.2305.10721 |
| [28] |
Z. P. Cao, Z. M. Sha, D. Lv, P. Z. Wei, B. W. Xiong, S. R. Ye, MGAMNet: a multi-granularity aware and hierarchically mixed network for BDS-3 satellite clock bias prediction, GPS Solut., 30 (2026), 13. https://doi.org/10.1007/s10291-025-01984-9 doi: 10.1007/s10291-025-01984-9
|
| [29] | D. Hendrycks, K. Gimpel, Gaussian error linear units (GELUs), arXiv preprint arXiv: 1606.08415, 2023. https://doi.org/10.48550/arXiv.1606.08415 |
| [30] | A. L. Zeng, M. X. Chen, L. Zhang, Q. Xu, Are transformers effective for time series forecasting? arXiv preprint arXiv: 2205.13504, 2022. https://doi.org/10.48550/arXiv.2205.13504 |
| [31] |
C. S. Fiskin, O. Turgut, S. Westgaard, A. G. Cerit, Time series forecasting of domestic shipping market: comparison of SARIMAX, ANN-based models and SARIMAX-ANN hybrid model, Int. J. Shipp. Transp. Logist., 14 (2022), 193-221. https://doi.org/10.1504/IJSTL.2022.122409 doi: 10.1504/IJSTL.2022.122409
|
| [32] |
C. C. Hu, Cargo volume forecast and personnel scheduling model of logistics network based on ARIMA, LSTM and MOP, ICEACE, 4 (2024), 1250-1254. https://doi.org/10.1109/ICEACE63551.2024.10898553 doi: 10.1109/ICEACE63551.2024.10898553
|
| [33] | H. Y. Zhou, S. H. Zhang, J. Q. Peng, S. Zhang, J. Li, H. Xiong, et al., Informer: Beyond efficient transformer for long sequence time-series forecasting, arXiv preprint arXiv: 2012.07436, 2021. https://doi.org/10.48550/arXiv.2012.07436 |
| [34] | Y. Q. Nie, N. H. Nguyen, P. Sinthong, J. Kalagnanam, A time series is worth 64 words: Long-term forecasting with transformers, arXiv preprint arXiv: 2211.14730, 2023. https://doi.org/10.48550/arXiv.2211.14730 |
| [35] | S. J. Bai, J. Z. Kolter, V. Koltun, An empirical evaluation of generic convolutional and recurrent networks for sequence modeling, arXiv preprint arXiv: 1803.01271, 2018. https://doi.org/10.48550/arXiv.1803.01271 |
Figures(11) / Tables(4)
Xiaoyuan Cheng. ACVPNet: An air cargo volume prediction model for China based on periodic structure and multi-dimensional feature interaction[J]. Journal of Industrial and Management Optimization, 2026, 22(5): 2428-2450. doi: 10.3934/jimo.2026089
DownLoad: