Research On Shared Bicycle System Analysis And Traffic Forecasting Methods

The shared bicycle system is a new type of transportation that has emerged in recent years.More and more people at home and abroad will choose to ride a shared bicycle.Although sharing bicycles brings a lot of convenience to our lives,there are still some problems.Since the rental and return of shared bicycles at different sites at different times are still unbalanced,this balance needs to be constantly updated.We solve this problem by forecasting and reallocating.In this paper,we mainly study the problem of site traffic prediction in the shared bicycle system.The main work of this paper is summarized as follows:1.A hierarchical prediction model of shared bicycle system based on Affinity Propagation(AP)clustering is proposed.Firstly,a two-layer neighborhood propagation clustering algorithm(TL-AP)is proposed.The TL-AP clustering algorithm is used to classify shared bicycle stations into several categories,which take into account the geographical location information of shared bicycle stations and the migration trend between stations.information.Compared with other clustering algorithms such as K-means,TL-AP clustering does not need to set the initial cluster center and the number of clusters(K value)in advance,which reduces the error caused by subjective experience to some extent.Then use the multi-similar reference model to predict the migration ratio between the internal sites,classes and classes of the class,combined with the predicted leased traffic of all the sites,so as to derive the shared bicycle rental and each class(station)Return the situation.This paper uses the New York(NYC)shared bicycle system to verify the proposed prediction model and compare it with the results of some widely used prediction models.The experimental results show that the proposed prediction model is better than those methods..2.A hierarchical prediction model of shared bicycle system based on trend iterative Gaussian Mixture Model(GMM)clustering is proposed.Firstly,a GMM clustering algorithm based on trend iteration(TL-GMM)is proposed.The TL-GMM clustering algorithm divides the shared bicycle station into multiple classes.The geographical location information of the site and the inter-site are considered in the cluster.Migrate trend information,use the paradigm to reduce the trend and perform cluster iteration.Compared with other clustering algorithms such as K-means,TL-GMM clustering does not get a certain clustering marker when projecting to data sample points during clustering,but obtains the probability of each class.The class has an important meaning and can improve the clustering accuracy.Second,use the Gradient Boosting Regressor Tree(GBRT)to predict the entire leased out traffic.Finally,the multi-similar reference model is used to predict the ratio of renting and returning between shared bicycles,so as to derive the migration of shared bicycles in each class(station)in the future.In this paper,the proposed prediction models are validated on shared bicycle systems in New York City(NYC)and Washington,DC(DC),and the prediction results and some currently widely used prediction methods and previously proposed two-layer neighborhood propagation clustering.Compared with the prediction results of the shared bicycle system hierarchical prediction model,the experimental results show that the proposed prediction model is superior to these methods.