3D convolutional neural network traffic data prediction method and model based on dimension protection decomposition

The invention discloses a 3D convolutional neural network traffic data prediction method and model based on dimension protection decomposition, a new dimension protection decomposition method is used to extract space-time characteristics in missing data, and then a 3D convolutional neural network is used to learn space-time dynamic characteristics in extracted potential characteristics. The model provided by the invention provides a new solution thought for a traffic data prediction problem, that is, dimension protection decomposition is used as a feature extractor of the traffic data prediction model, and then the features extracted by the dimension protection decomposition are subjected to space-time dynamic feature modeling through a 3D convolutional neural network. Compared with other complementation methods, the dimension protection decomposition method is better in performance and higher in space-time characteristic extraction capability. In addition, the 3D convolutional neural network is adopted, so t