Research On Traffic Flow Prediction Method Based On Spatiotemporal Clustering And Gaussian Nearest Neighbor

With the rapid development of China’s economy,cars are becoming more and more popular in Chinese families,which also brings about the problem of traffic congestion.In particular,in some large and medium-sized cities,urban traffic congestion intensified year by year,the congestion lasted for several hours,causing a huge loss of time.As the end of the city,the logistics and distribution link is also the most critical link in the distribution link,whether the goods can be delivered to the customer in time will directly affect the customer’s satisfaction with the distribution service.The serious traffic congestion in cities has greatly reduced the distribution efficiency.In the process of distribution,how to avoid road congestion or choose appropriate distribution routes when road congestion is not clear,so as to complete the distribution on time has become the target pursued by the distribution industry.In order to quantitatively evaluate the state of traffic congestion,many cities in China have developed their own congestion index system,which can reflect the running situation of road traffic in real time through the congestion index.Meanwhile,it also provides the prediction of future traffic congestion index,so as to provide decision support for the distribution service of logistics enterprises.However,the existing traffic congestion index prediction system can only predict a short period of time,the accuracy of the prediction is not high,can not predict the future period of traffic status.In order to predict the traffic flow over a period of time in the future,this article establishes a traffic flow prediction model based on the traffic congestion index,and proposes a two-step heuristic algorithm to predict the traffic flow by using Gaussian neighbor in the cluster after the spatial and temporal clustering.Then,the article takes the traffic congestion index in the main urban area of Shanghai as an example to verify the validity of the prediction model and algorithm.Articles on the time dimension is established first k-means clustering model,established on the spatial dimension model of DBSCAN clustering based on density,for two weeks in Shanghai in 68 area traffic congestion index data in time and space clustering analysis,obtained the Shanghai regional traffic model of nine kinds of clustering results,and k neighbor method is used to cluster the data set to update the search.In the updated cluster,the traffic congestion index under nine modes is respectively predicted by using gaussian neighbor method,and the change rule of traffic congestion index in the next week is obtained.Moreover,the prediction results of cyclic neural network algorithm(CNN)and long-term memory neural network(LSTM)are compared and analyzed.Studies show that gaussian neighbor prediction method based on spatial and temporal clustering,because in time and space,on the basis of clustering,clustering of data sets with the same characteristics of the data set after the complete update k neighbor search,less noise,using gaussian neighbor prediction algorithm.It can more accurately reflect the relationship between neighboring points,thus the cycle of neural network algorithm memory neural network(CNN)and length(LSTM)method of the prediction results more accurate,more fitting.