Study and Application of Multi Fine-grained Tunnel Traffic Flow Prediction Model Based on Spatio-temporal Relation
The tunnel traffic flow prediction is an important link in tunnel traffic early warning and control.In practice,the different fine-grained traffic flow predictions are usually performed based on the multiple variables with spatio-temporal relation.Therefore,it is challenging to use multi-source variables to accurately predict the fine-grained traffic flows at different times.It is usually to capture and fit time series mapping relation with the existing methods based on the single-source time variables,or improve the feature extraction capabilities in multi-source sequence data for single-scale prediction.These methods lack the ability to capture spatio-temporal relation among multi-source variables,and there are deficiencies in multivariate processing and fine-grained feature extraction.That leads to an inability to fully address the spatio-temporal relation in complex traffic environments,resulting in poor prediction performance.To solve these problems,based on the spatio-temporal relation,the multi fine-grained tunnel traffic flow prediction model,including multi fine-grained fusion prediction module and spatio-temporal relation feature extraction module,was proposed.The multi fine-grained fusion prediction module was used to fuse multiple variables,and extract the fine-grained features at different time scales,thus ensuring the model adapt to complex and changing traffic environments.Subsequently,the spatio-temporal relation feature extraction module further processed these features to capture the spatio-temporal dependencies among variables,thereby achieving the accurately prediction on traffic flow trends.The proposed model combined the advantages of spatio-temporal relation and multi fine-grained feature extraction.It can effectively meet the needs of traffic flow prediction at different time scales.The model was compared in 3 traffic flow data sets(PEMS4,PEMS8,and China tunnel traffic flow data).The result indicates that the proposed method performs well in the time series prediction tasks,especially in multi fine-grained tunnel traffic flow prediction.
NETL
NSTL
万方数据
