Short-term Traffic Flow Prediction Method Considering Information Security for Connected Vehicles

Abstract:

Aiming at the information security and data missing caused by network attacks or interference for ICVs （Intelligent Connected Vehicle）， this paper proposes a short-term prediction method of traffic flow state based on data imputation. Firstly， based on the edge computing task offloading model， the abnormal data of the V2X （vehicle to everything， V2X） communication process of ICVs is identified dynamically. Secondly， a graph embedding （GE） with data imputation mechanism and long short-term memory （LSTM） neural network model is proposed to impute the missing data. Thirdly， the neural network model is established based on the complete data sets to realize the short-term traffic flow state prediction. Finally， by the proposed model applied in Beijing for field test， the final results show that the short-term prediction effect of traffic flow state is significantly improved. In comparison with other methods， the prediction error is reduced by 87.4% and the accuracy of the prediction effect is 95% according to the actual traffic flow state， which provides a novel theoretical support and technical solution for vehicular information security and dynamic optimization of traffic resources in the intelligent networked environment.