Spatio-temporal Co-location Congestion Pattern Mining Based On Clustering

Nowadays,traffic pressure is increasing,and traffic congestion is becoming more and more serious everywhere.Especially when people meet the rush hour,they may stay for half an hour or more at a certain section or intersection,which leads to a lot of time being wasted.As a transportation hub,urban intersections bear tremendous traffic pressure.Using spatial juxtaposition pattern mining algorithm to process traffic data can discover juxtaposition patterns in urban traffic congestion,thus providing decision support for traffic managers and alleviating traffic pressure.However,the traditional spatial juxtaposition pattern mining algorithm does not take time attributes into account.Moreover,because the proximity measure chosen by the traditional spatial juxtaposition pattern mining method is mostly the distance between instances,the result can only see some local correlation in urban traffic,and can not be observed from the perspective of the whole urban traffic network.In addition,for traffic flow data,there is no clear criterion to judge when the traffic value of the section is already in the congestion range,so as to judge that the section is in the congestion state.Therefore,the average method used in our daily life is not suitable for dealing with traffic flow data at intersections.Therefore,this paper will study from two main aspects,how to judge the congestion situation of the intersection through traffic flow data,how to mine spatiotemporal juxtaposition congestion patterns with time attributes,and provide a spatiotemporal juxtaposition congestion pattern mining prototype system for usersFor the first aspect,this paper proposes a corresponding clustering algorithm to pre-process traffic flow data at intersections,so as to judge the congestion status through traffic flow data at intersections.Clustering algorithm can classify unlabeled data according to their properties,thus classifying data with similar properties.Therefore,by clustering,we can define a more appropriate congestion judgment condition,so as to identify congestion instances.For the second aspect,we change the traditional proximity relation of instances and take the traffic value as the judgment condition.At the same time,we mine different instances in different time periods,so as to take the time attributes into account,and propose a spatiotermporal parallel congestion pattern mining algorithm.In addition,this paper uses real data to verify the correctness and effectiveness of the algorithm.In this paper,the implementation process of a spatiotemporal congestion pattern mining prototype system is introduced in detail.By inputting real data,the prototype system can obtain effective spatio-temporal congestion patterns based on spatio-temporal congestion pattern rmining algorithm,and visually display them to users,providing a reference for users to travel,so that users can choose an appropriate route to avoid congestion.At the same time,it can also help traffic managers to formulate traffic strategies at intersections,thereby alleviating traffic congestion.