Design And Implementation Of Efficient Clustering Algorithm For Large Amounts Of Trajectory Data

With the maturity of mobile Internet and sensor technology,it is more convenient for people to obtain real-time location data.As a result,it becomes easier and easier to obtain trajectory data of moving objects,thus accumulating a large amount of trajectory data.These trajectory data contain a lot of valuable information that reflects the behavior of moving objects.Clustering is a common and basic method for mining trajectory data.It can find trajectories with the same or similar characteristics,laying a foundation for applications such as trajectory anomaly detection and intelligent traffic prediction.In this paper,an in-depth study of the efficient clustering method that integrates multiple characteristics of the trajectory,fully considers the spatiotemporal characteristics and motion attributes of the trajectory data,and combines K-Means,Local Outlier Factor(LOF),and Chameleon.With the support of the Map Reduce framework,we designed and implemented a parallel and incremental clustering algorithm for large-scale vehicle trajectory data.While improving the accuracy of clustering analysis of trajectory data,it greatly improved the clustering of large-scale trajectory data.effectiveness.The main research content of this article includes the following three aspects:(1)Aiming at the problem that the trajectory similarity measurement cannot fully reflect the trajectory characteristics caused by incomplete consideration of trajectory attributes and insufficient fusion in the process of vehicle trajectory clustering,the FK-Means clustering algorithm based on trajectory multi-feature fusion is proposed.It uses the discrete Fréchet distance measurement method and the K-Means algorithm to combine multiple attributes of trajectories to calculate the similarity between trajectories,and solves the problem of vehicle trajectory clustering with multi-attribute characteristics.Experimental results show that the proposed method has a better effect on discovering common features of trajectory data.(2)Aiming at the problem of low efficiency in cluster analysis of a large number of trajectory data,the FK-Means-LOF algorithm based on the Map Reduce framework is proposed.The algorithm first uses the Map Reduce framework to realize the parallel processing of the FK-Means algorithm,and then uses the improvement The LOF algorithm optimizes the results obtained by parallel clustering,and improves the problem of all clustering accuracy drops caused by the parallelization process.Experimental results show that the algorithm effectively improves the clustering efficiency of large-scale trajectory data under the premise of ensuring the global accuracy of the trajectory clustering process.(3)Aiming at the problem that the trajectory data is updated quickly and the traditional clustering algorithm cannot locally re-cluster efficiently,a model for incremental clustering of trajectory data and its implementation algorithm FK-Means-DC are proposed.The algorithm is mainly divided into two In the first stage,the FK-means algorithm is used to cluster the trajectory data,and the second stage is to combine the chameleon algorithm and the double breadth first search(DBFS,Double breadth First Search)algorithm to integrate the clustering results and avoid a small amount of new additions All trajectory data re-clustering consumption caused by trajectory improves the efficiency of trajectory clustering analysis.Experimental results show that the proposed local incremental clustering algorithm has good generalization ability,which is helpful for the application of delay-sensitive clustering analysis projects.