隨著深度學習的持續發展, 研究人員會利用都市中的感測器系統將人類的移動資訊數據化並建立預測模型來幫助如都市規劃, 旅遊規劃, 大眾運輸調配, 商場活動等等的相關應用.然而近年因為COVID-19疫情爆發, 都市中人群的移動模式產生劇烈變動, 導致先前訓練好的模型無法準確預測發生變化的全市人流量.因此為了準確預測全市的人流並且能夠根據發生變化的人流更新預測模型, 本研究旨在使用全市人流圖像預測網路CFRBF-UNet進行知識遷移來準確預測全市的人流.我們的模型透過知識轉移的能力能夠僅使用少量的訓練資料更新模型, 同時我們的模型能夠有效萃取人流的時空依賴性以及考量外部因素對人流的影響, 使其快速恢復原來的預測性能繼續使用.本研究的實驗使用台灣真實資料來驗證所提出方法的有效性.與現有的方法相比, 我們的模型擁有最好的預測性能, 並且我們也為CFRBF-UNet的內部元件進行廣泛的實驗來評估各元件對於預測性能的貢獻.最後我們也評估了所提出的方法在遷移學習時的轉移速度以及精準度的有效性, 實驗表明我們的模型能夠僅使用1%的參數進行Fine-tuning, 其節省了62%的訓練時間來快速使模型快速恢復原來的預測性能。

As deep learning advances, sensor systems for smart cities progress. These systems digitize information on human movement to build predictive models for urban planning, tourism planning, mass transportation deployment, and shopping mall layouts, among others. However, the COVID-19 outbreak caused drastic changes in the movement patterns of urban populations, reducing the accuracy of previously-trained models for the prediction of city-wide crowd flow. The current paper thus updates prediction models using the CFRBF-UNet with knowledge transfer learning. Through knowledge transfer, our model is able to update using only a small amount of training data. The proposed model successfully extracts the spatial and temporal dependence of crowd flow and considers the impact of external factors, thereby quickly restoring the original prediction performance. We used real data from Taiwan to verify the effectiveness of the proposed method, proving its superiority over existing methods in terms of prediction performance. We also conducted extensive experiments on the internal components of the CFRBF-UNet to evaluate the contribution of each component. Finally, we evaluated the effectiveness of the proposed method in terms of transfer speed and accuracy during transfer learning. The experiments show that the proposed model is able to fine-tune using only 1% of the parameters, which reduces training time by 62%.

[1] L. Bai, L. Yao, C. Li, X. Wang, and C. Wang, "Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting," arXiv preprint arXiv:2007.02842, 2020.
[2] D. Chen, "Research on traffic flow prediction in the big data environment based on the improved RBF neural network," IEEE Transactions on Industrial Informatics, vol. 13, no. 4, pp. 2000-2008, 2017.
[3] J. Chen et al., "Fine-grained prediction of urban population using mobile phone location data," International Journal of Geographical Information Science, vol. 32, no. 9, pp. 1770-1786, 2018.
[4] CWB Observation Data Inquire System v7.2. [Online]. Available: https://e-service.cwb.gov.tw/HistoryDataQuery/index.jsp. [Accessed: 11-May-2022].
[5] Y.-C. Chen and D.-C. Li, "Selection of key features for PM2.5 prediction using a wavelet model and RBF-LSTM," Applied Intelligence, vol. 51, no. 4, pp. 2534-2555, 2021.
[6] C. Ding, J. Duan, Y. Zhang, X. Wu, and G. Yu, "Using an ARIMA-GARCH Modeling Approach to Improve Subway Short-Term Ridership Forecasting Accounting for Dynamic Volatility," IEEE Transactions on Intelligent Transportation Systems, vol. 19, no. 4, pp. 1054-1064, 2018.
[7] G. Du, X. Cao, J. Liang, X. Chen, and Y. Zhan, "Medical image segmentation based on u-net: A review," Journal of Imaging Science and Technology, vol. 64, no. 2, pp. 20508-1-20508-12, 2020.
[8] R. Fu, Z. Zhang, and L. Li, "Using LSTM and GRU neural network methods for traffic flow prediction," in 2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC), 2016: IEEE, pp. 324-328.
[9] A. Ganjavi, E. Christopher, C. M. Johnson, and J. Clare, "A study on probability of distribution loads based on expectation maximization algorithm," in 2017 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT), 2017: IEEE, pp. 1-5.
[10] Vidnerová, Petra, “RBF-Keras: an RBF Layer for Keras Library,” GitHub. [Online]. Available: https://github.com/PetraVidnerova/rbf_keras. [Accessed: 30-Jul-2022].
[11] A. P. Gopi, R. N. S. Jyothi, V. L. Narayana, and K. S. Sandeep, "Classification of tweets data based on polarity using improved RBF kernel of SVM," International Journal of Information Technology, pp. 1-16, 2020.
[12] Google, “Covid-19 Community mobility reports.” [Online]. Available: https://www.google.com/covid19/mobility/. [Accessed: 27-May-2022].
[13] I. H. Sarker, "A machine learning based robust prediction model for real-life mobile phone data," Internet of Things, vol. 5, pp. 180-193, 2019.
[14] J. Hu, L. Shen, and G. Sun, "Squeeze-and-excitation networks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132-7141.
[15] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[16] Q. Hu, R. Zhang, and Y. Zhou, "Transfer learning for short-term wind speed prediction with deep neural networks," Renewable Energy, vol. 85, pp. 83-95, 2016.
[17] S. Han, C. Qubo, and H. Meng, "Parameter selection in SVM with RBF kernel function," in World Automation Congress 2012, 2012: IEEE, pp. 1-4.
[18] S. Hao, M. Zhang, and A. Hou, "Short-term Traffic Flow Forecast Based on DE-RBF Fussion Model," in Journal of Physics: Conference Series, 2021, vol. 1910, no. 1: IOP Publishing, p. 012035.
[19] Y. Han, S. Wang, Y. Ren, C. Wang, P. Gao, and G. Chen, "Predicting station-level short-term passenger flow in a citywide metro network using spatiotemporal graph convolutional neural networks," ISPRS International Journal of Geo-Information, vol. 8, no. 6, p. 243, 2019.
[20] Y.-L. He, X.-Z. Wang, and J. Z. Huang, "Fuzzy nonlinear regression analysis using a random weight network," Information Sciences, vol. 364, pp. 222-240, 2016.
[21] Z. Jiang, P. Tahmasebi, and Z. Mao, "Deep residual U-net convolution neural networks with autoregressive strategy for fluid flow predictions in large-scale geosystems," Advances in Water Resources, vol. 150, p. 103878, 2021.
[22] Editor Group, K., "Asia Taiwan Taipei 4 Days Travel Guide," KKday Blog, 21-Nov-2017. [Online]. Available: https://blog.kkday.com/8088/asia-taiwan-taipei-4-days-travel-guide. [Accessed: 10-May-2022].
[23] A. Lin, B. Chen, J. Xu, Z. Zhang, G. Lu, and D. Zhang, "Ds-transunet: Dual swin transformer u-net for medical image segmentation," IEEE Transactions on Instrumentation and Measurement, 2022.
[24] G. Li, Q. Zhang, Q. Lin, and W. Gao, "A three-level radial basis function method for expensive optimization," IEEE Transactions on Cybernetics, 2021.
[25] K.-L. Li, C.-J. Zhai, and J.-M. Xu, "Short-term traffic flow prediction using a methodology based on ARIMA and RBF-ANN," in 2017 Chinese Automation Congress (CAC), 2017: IEEE, pp. 2804-2807.
[26] S. Y. Liu, S. Liu, Y. Tian, Q. L. Sun, and Y. Y. Tang, "Research on Forecast of Rail Traffic Flow Based on ARIMA Model," in Journal of Physics: Conference Series, 2021, vol. 1792, no. 1: IOP Publishing, p. 012065.
[27] T. Li, J. Zhang, K. Bao, Y. Liang, Y. Li, and Y. Zheng, "Autost: Efficient neural architecture search for spatio-temporal prediction," in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2020, pp. 794-802.
[28] Y. Liang et al., "Fine-Grained Urban Flow Prediction," in Proceedings of the Web Conference 2021, 2021, pp. 1833-1845.
[29] Y. Liu, H. Zheng, X. Feng, and Z. Chen, "Short-term traffic flow prediction with Conv-LSTM," in 2017 9th International Conference on Wireless Communications and Signal Processing (WCSP), 2017: IEEE, pp. 1-6.
[30] Y. Lu, Z. Tian, P. Peng, J. Niu, W. Li, and H. Zhang, "GMM clustering for heating load patterns in-depth identification and prediction model accuracy improvement of district heating system," Energy and Buildings, vol. 190, pp. 49-60, 2019.
[31] Y. Lv, Y. Duan, W. Kang, Z. Li, and F.-Y. Wang, "Traffic flow prediction with big data: a deep learning approach," IEEE Transactions on Intelligent Transportation Systems, vol. 16, no. 2, pp. 865-873, 2014.
[32] Z. Lin, J. Feng, Z. Lu, Y. Li, and D. Jin, "Deepstn+: Context-aware spatial-temporal neural network for crowd flow prediction in metropolis," in Proceedings of the AAAI Conference on Artificial Intelligence, 2019, vol. 33, no. 01, pp. 1020-1027.
[33] A. E. Orhan and X. Pitkow, "Skip connections eliminate singularities," arXiv preprint arXiv:1701.09175, 2017.
[34] O. Oktay et al., "Attention u-net: Learning where to look for the pancreas," arXiv preprint arXiv:1804.03999, 2018.
[35] N. S. Punn and S. Agarwal, "Modality specific U-Net variants for biomedical image segmentation: a survey," Artificial Intelligence Review, pp. 1-45, 2022.
[36] S. Pineda-Arango, D. Obando-Paniagua, A. Dedeoglu, P. Kurzendörfer, F. Schestag, and R. Scholz, "Improving Sample Efficiency with Normalized RBF Kernels," arXiv preprint arXiv:2007.15397, 2020.
[37] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," in International Conference on Medical image computing and computer-assisted intervention, 2015: Springer, pp. 234-241.
[38] C. Sun, M. Ma, Z. Zhao, S. Tian, R. Yan, and X. Chen, "Deep transfer learning based on sparse autoencoder for remaining useful life prediction of tool in manufacturing," IEEE Transactions on Industrial Informatics, vol. 15, no. 4, pp. 2416-2425, 2018.
[39] D. Stoller, S. Ewert, and S. Dixon, "Wave-u-net: A multi-scale neural network for end-to-end audio source separation," arXiv preprint arXiv:1806.03185, 2018.
[40] H. Shi, L. Chen, Z. Xu, and D. Lyu, "Personalized location recommendation using mobile phone usage information," Applied intelligence, vol. 49, no. 10, pp. 3694-3707, 2019.
[41] J. Long, E. Shelhamer, and T. Darrell, "Fully convolutional networks for semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3431-3440.
[42] J. Schmidt-Hieber, "Nonparametric regression using deep neural networks with ReLU activation function," The Annals of Statistics, vol. 48, no. 4, pp. 1875-1897, 2020.
[43] S. Sun, H. Wu, and L. Xiang, "City-wide traffic flow forecasting using a deep convolutional neural network," Sensors, vol. 20, no. 2, p. 421, 2020.
[44] Z. Shen, Y. Xu, B. Ni, M. Wang, J. Hu, and X. Yang, "Crowd counting via adversarial cross-scale consistency pursuit," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 5245-5254.
[45] C. Tian, X. Zhu, Z. Hu, and J. Ma, "Deep spatial-temporal networks for crowd flows prediction by dilated convolutions and region-shifting attention mechanism," Applied Intelligence, vol. 50, no. 10, pp. 3057-3070, 2020.
[46] Taiwan Centers for Disease Control, “CECC raises epidemic warning to level 3 nationwide from May 19 to May 28; strengthened measures and restrictions introduced across Taiwan to reduce community transmission.” [Online]. Available: https://www.cdc.gov.tw/En/Bulletin/Detail/VN_6yeoBTKhRKoSy2d0hJQ?typeid=158. [Accessed: 28-May-2022].
[47] V. K. Valloli and K. Mehta, "W-net: Reinforced u-net for density map estimation," arXiv preprint arXiv:1903.11249, 2019.
[48] Z. Wang, Z. Dai, B. Póczos, and J. Carbonell, "Characterizing and avoiding negative transfer," in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 11293-11302.
[49] J. Wainer and P. Fonseca, "How to tune the RBF SVM hyperparameters? An empirical evaluation of 18 search algorithms," Artificial Intelligence Review, vol. 54, no. 6, pp. 4771-4797, 2021.
[50] J. Wu, J. Long, and M. Liu, "Evolving RBF neural networks for rainfall prediction using hybrid particle swarm optimization and genetic algorithm," Neurocomputing, vol. 148, pp. 136-142, 2015.
[51] L. Wang, X. Geng, X. Ma, F. Liu, and Q. Yang, "Cross-city transfer learning for deep spatio-temporal prediction," arXiv preprint arXiv:1802.00386, 2018.
[52] S. Wang, H. Miao, J. Li, and J. Cao, "Spatio-Temporal Knowledge Transfer for Urban Crowd Flow Prediction via Deep Attentive Adaptation Networks," IEEE Transactions on Intelligent Transportation Systems, 2021.
[53] B. Yu, H. Yin, and Z. Zhu, "St-unet: A spatio-temporal u-network for graph-structured time series modeling," arXiv preprint arXiv:1903.05631, 2019.
[54] H. Yao, Y. Liu, Y. Wei, X. Tang, and Z. Li, "Learning from multiple cities: A meta-learning approach for spatial-temporal prediction," in The World Wide Web Conference, 2019, pp. 2181-2191.
[55] I. Yilmaz and O. Kaynar, "Multiple regression, ANN (RBF, MLP) and ANFIS models for prediction of swell potential of clayey soils," Expert systems with applications, vol. 38, no. 5, pp. 5958-5966, 2011.
[56] J. Zeng and W. Qiao, "Short-term solar power prediction using an RBF neural network," in 2011 IEEE power and energy society general meeting, 2011: IEEE, pp. 1-8.
[57] J. Zhang et al., "LCU-Net: A novel low-cost U-Net for environmental microorganism image segmentation," Pattern Recognition, vol. 115, p. 107885, 2021.
[58] J. Zhang, Y. Zheng, and D. Qi, "Deep spatio-temporal residual networks for citywide crowd flows prediction," in Thirty-first AAAI conference on artificial intelligence, 2017.
[59] J. Zhang, Y. Zheng, D. Qi, R. Li, and X. Yi, "DNN-based prediction model for spatio-temporal data," in Proceedings of the 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2016, pp. 1-4.
[60] Y. Zhou, J. Li, H. Chen, Y. Wu, J. Wu, and L. Chen, "A spatiotemporal hierarchical attention mechanism-based model for multi-step station-level crowd flow prediction," Information Sciences, vol. 544, pp. 308-324, 2021.