| With the continuous development of urbanization,the urban population and the number of cars in the world are increasing year by year,and the basic transportation construction in most cities can no longer meet people’s travel needs.Too many cars cause urban traffic to run overloaded,which is easy to cause traffic congestion.Traffic congestion not only wastes people’s travel time and increases unnecessary travel costs,but also greatly increases the amount of car exhaust emissions,resulting in increased atmospheric pollution and the greenhouse effect.Therefore,alleviating the occurrence of traffic congestion has become one of the problems that people urgently need to solve.Under this circumstance,how to classify traffic states accurately,determine traffic congestion levels based on traffic data,and further realize the prediction of traffic flow data has become an important research topic.In recent years,artificial intelligence has developed rapidly in various fields using its own huge advantages,and achieved good application results.Therefore,this paper uses artificial intelligence technology to carry out simulation experiments and algorithm improvement research on traffic congestion discrimination and traffic flow prediction.The core content of the thesis is divided into two parts: the first part,the improved algorithm based on the fuzzy C-means clustering algorithm to divide the traffic flow data clustering into different traffic states and determine the traffic congestion level;the second part,based on empirical mode decomposition The improved algorithm combined with recurrent neural network predicts traffic flow data.First,based on the analysis of the characteristics of traffic flow parameters,vehicle flow,road space occupancy rate and average vehicle speed are selected as the characteristic values of the traffic flow data in this paper.Preprocess the original traffic flow data.Then,the fuzzy C-means clustering algorithm is used to divide the traffic state according to the characteristics of the ambiguity of the traffic state.In order to improve the shortcomings of the algorithm’s instability and easy to fall into the local optimal solution,a FCM clustering algorithm based on cuckoo search algorithm and Canopy algorithm optimization is proposed to identify traffic jams.The improved algorithm can automatically determine the number of clusters,improve the local optimal solution phenomenon,and has better convergence effect and higher stability.Finally,an algorithm combining empirical mode decomposition and recurrent neural network is used to predict traffic flow data.This part first introduces the structure of the recurrent neural network,and its two variants of long-term and short-term memory and gated recurrent unit.Then,combined with empirical mode decomposition,a traffic flow prediction model based on spatio-temporal characteristics is proposed.After comparing the error performance indexes of several algorithms,the result shows that the EMD-GRU algorithm has better prediction effect.Based on the EMD-GRU algorithm,a modified ensemble empirical mode decomposition(MEEMD)is proposed to further improve the accuracy of traffic flow prediction. |
