To implement the congestion control goals outlined in the Outline for Building a Leading Transportation Nation and address traffic congestion during urban peak hours, this paper integrates theories of multimodal information fusion and time-series analysis to propose a multimodal data fusion-based traffic flow prediction model for peak-hour scenarios. The model centers on minute-level traffic flow data from four key intersections in urban multi-functional zones, incorporating multidimensional multimodal information. Through data preprocessing, core features of traffic flow are extracted. Six mainstream prediction models are selected to establish a comparative framework, with quantitative evaluation based on MAE, RMSE, and R². Model weights are determined using the entropy weight method to construct a weighted fusion model. Validated with actual data from the 2016–2017 period, the fusion model outperforms all individual models across all metrics: MAE and RMSE are reduced to 16.4 and 22.3, respectively, while R² improves to 0.68. The correlation coefficient with actual traffic flow reaches 0.89, and trend consistency during critical periods achieves 92%, enabling accurate capture of dynamic traffic patterns. This model provides scientific support for traffic management and resident travel, while also offering a new technical pathway for the development of intelligent transportation systems.

Luo Shunan, Sun Xiaoqing, Deng Yun, Zhang Yucheng. Research on Multimodal Data Fusion for Traffic Flow Prediction during Peak Hours. Academic Journal of Computing & Information Science (2026), Vol. 9, Issue 2: 61-66. https://doi.org/10.25236/AJCIS.2026.090209.

[1] Liu Y ,Jia C ,Rasouli S , et al.SAFER-predictor: Sparse adversarial training framework for robust traffic prediction under missing and noisy data[J].Communications in Transportation Research, 2025,5:100192.

[2] Zhang H ,Yi M ,Zhang X , et al.Multi-scale dual dynamic spatiotemporal graph convolutional network for traffic flow prediction[J].Applied Intelligence,2025,55(12):828.

[3] Li W ,Yang G ,Xiong Z, et al.A Traffic Flow Prediction Model Based on Dynamic Graph Convolution and Adaptive Spatial Feature Extraction[J].Symmetry,2025,17(7):1007.

[4] Sun H ,Li P,Li Y .Traffic Flow Prediction in Data-Scarce Regions: A Transfer Learning Approach[J].Computers, Materials & Continua,2025,83(3):4899-4914.

[5] Kumar R ,Moreira M J ,Chandra J .CSCN: an efficient snapshot ensemble learning based sparse transformer model for long-range spatial-temporal traffic flow prediction[J].Data Mining and Knowledge Discovery,2025,39(4):31.

[6] M K ,K J ,K A .Multivariate N-bEATS Algorithm for Vehicular Traffic Flow Prediction Analysis For Indian Road Scenario[J].International Journal of Intelligent Transportation Systems Research, 2025,23(2):1-15.

[7] Zhang C,You Y,Pan G,et al.Real-time traffic turning flow forecasting under few-shot and zero-shot scenarios by decoder-only model[J].Artificial Intelligence for Transportation,2026,5:100046.

[8] Qiu Z,Wu H,Teng G,et al.Multi-scale spatial-temporal transformer for traffic flow prediction.[J]. Scientific reports,2025,16(1):3531.

[9] Li W,Wei H ,Liu X , et al.Enhancing traffic flow prediction through multi-view attention mechanism and dilated convolutional networks[J].Complex & Intelligent Systems,2025,12(1):37.

[10] Chu Y ,Fu T ,Liu P , et al.STICformer: spatio-temporal intrinsic connections transformer for traffic flow prediction.[J].Scientific reports,2025,16(1):1881.