Multi-Step Prediction Technology for Road PCI Based on Adaptive Road Map Expression

Based on the shortcomings of existing PCI prediction technologies, and fully considering the spatial latitude and temporal latitude characteristics of roads with adaptive graph convolutional neural network technology, we propose an adaptive road map embedded PCI multi-step prediction technology based on Transformer. And it is verified by practical cases and compared with various model prediction methods. The accurate prediction of road PCI is realized. The results show that the PCI prediction technology based on the road segment adaptive diagram uses a deep learning method to realize multi-step prediction of road PCI, which has the characteristics of high precision and high robustness. The RMSE and MAPE of this method are the smallest among common prediction methods, indicating that this model can accurately predict PCI with high prediction accuracy and can provide accurate road PCI data support for urban planning, traffic management, and other related fields.