Study Of Short-term Traffic Flow Forecasting Based On Flower Pollination Algorithm And Neural Network

With the rapid development of China's economy and the accelerated increase of urbanization level,the number of vehicles continues to grow,which has led to a series of traffic problems such as traffic congestion,traffic accidents and air pollution.Among them,people are paying more and more attention to traffic congestion.In order to solve this problem,an Intelligent Transportation System(ITS)was created.It is an effective method to solve urban traffic problems at present.It can induce and control traffic information in real time on the basis of existing technology,and maximize the transportation and management efficiency of transportation system.The key to real-time induction is to know the road traffic status in advance,which often requires accurate traffic flow prediction.Therefore,how to accurately predict short-term traffic flow is the key to ensure the effective operation of the intelligent transportation system.This paper originated from the pre-research project of the traffic guidance module of Hisense Network Technology Co.,Ltd.,Traffic Intelligent Control and Control Platform,focusing on short-term traffic flow prediction based on flower pollination algorithm and neural network at urban intersections.The research in this paper will be carried out in the following aspects:Firstly,it introduces the development of intelligent transportation system,expounds the basic principles of traffic flow prediction,and lays a theoretical foundation for the establishment of subsequent prediction models;Secondly,aiming at the shortcomings of genetic algorithm,a probabilistic adaptive genetic algorithm is proposed and three prediction models are established:traffic flow prediction model based on BP neural network,genetic neural network and probability adaptive genetic neural network.Through simulation and comparison experiments,it is concluded that the probability adaptive genetic neural network traffic flow prediction model has improved network performance compared with the other two models,and has achieved better prediction results;Then,in order to better solve the defect that the BP neural network has a slow convergence rate and is easy to fall into the local minimumvalue when adjusting the weight threshold,this paper chooses to optimize the Flower Pollination Algorithm(FPA)with better global optimization ability.BP neural network,in view of its own slow convergence rate,this paper improves the FPA algorithm from three aspects: the initial population optimization,the improved conversion probability and the distance difference local pollination strategy.Finally,through comparison with the IGA-WNN model,the improved FPA-BP neural network improves network performance,has faster convergence speed and higher learning and generalization ability,and can be used for traffic flow prediction of actual systems.