Research On Hybrid Prediction Algorithm Of SVM And BPNN Based On Road Occupancy

With the continuous development of our country's economy,the rapid progress of our society,and the improvement of our people's living standard,the private car ownership is rising year by year.Today,although various transportations make us feel fast and convenient,it brings us a series of problems,such as traffic congestion,environmental pollution,and traffic accidents,etc.In order to solve these problems,intelligent transportation systems are developed gradually,and the short-term traffic flow prediction becomes an important research topic of intelligent transportation systems.When the traffic flow is mentioned,the road occupancy has to be concerned.The level of the road occupancy reflects the traffic flow size.Using some efficient models to predict various traffic flows may alleviate traffic pressure,reduce environmental pollution,and facilitate citizens' transportation to some extent.Therefore,studying traffic flow prediction is of great significance for all aspects of our society.In the prediction of some traffic flows,support vector machine(SVM)is generally adopted in the presence of small amount of traffic data because SVM has the stronger prediction ability for the situation.However,the predictive performance will decrease as the amount of traffic data increases,thus using only SVM to predict traffic flow will inevitably cause a large error for large amount of traffic data.To overcome the shortcoming of SVM,We use BP neural network(BPNN)for short-term traffic flow prediction.In the meanwhile,BP neural network also has obvious disadvantages.For instance,although BP neural network can train the model on a huge amount of data,it can easily fall into a local optimum with a slow speed of the model training.This paper proposes a hybrid model based on SVM and BP neural network,which eliminate the shortcomings of the two single models andcombines the advantages of them in short-term traffic flow prediction.The improved Particle Swarm Optimization(PSO)algorithm is used in a hybrid model to optimize the weights.Simulation results demonstrate that the proposed hybrid model improved multiple performance indicators compared to SVM and BP neural network.