The invention provides a traffic flow prediction method based on a future knowledge differential transformation neural network. The method comprises the following steps: processing given traffic flow sample data into embedded data containing future statistics by adopting an embedder; respectively sensing the time-varying trend of the traffic flow in the embedded data and mining the global and local space-time dependency relationship of the traffic flow by an encoder through a high-order differential neural network and a double-space-time convolution module, and encoding the embedded data into hidden features; decoding the hidden features through a decoder in combination with progressive unbiased future statistics to obtain an initial prediction value of the traffic flow, a discrimination feature and an estimation value of the unbiased future statistics; and based on the discriminant features, adaptively fusing the preliminary predicted value and the estimated value of the unbiased future statistical magnitude through a gating technology, and outputting a final prediction result. According to the invention, data-driven perception of the time domain can be realized, and the accuracy and robustness of traffic flow prediction are effectively improved.
本发明提供一种基于未来知识差分变换神经网络的交通流量预测方法,包括:采用嵌入器将给定的交通流量样本数据处理为包含未来统计量的嵌入数据;由编码器通过高阶差分神经网络和双时空卷积模块,分别感知所述嵌入数据中交通流量的时变趋势和挖掘交通流量的全局和局部时空依赖关系,将所述嵌入数据编码为隐藏特征;通过解码器结合渐进无偏未来统计量对所述隐藏特征进行解码,得到交通流量的初步预测值、判别特征和无偏未来统计量的估计值;基于所述判别特征,通过门控技术自适应融合所述初步预测值和无偏未来统计量的估计值,输出最终预测结果。本发明能够对时域实现数据驱动式感知,有效提高了交通流量预测的准确性和鲁棒性。
Traffic flow prediction method based on future knowledge differential transformation neural network
一种基于未来知识差分变换神经网络的交通流量预测方法
2024-03-29
Patent
Elektronische Ressource
Chinesisch
Traffic flow prediction method based on neural network
Europäisches Patentamt | 2020
|Traffic flow prediction using neural network
IEEE | 2018
|Europäisches Patentamt | 2024
|