Trajectory Clustering And Anomaly Detection Algorithm For GPS Data

At present,with the extensive development of 5G communication technology and smart phones,people's life movement trajectories are collected and recorded,and some trajectory data sets are formed.How to use the analysis of these human mobile data to find valuable trajectory behavior information from a large amount of information,such as identifying abnormal trajectories,is the focus of current trajectory data research.Now,trajectory data research mainly includes trajectory clustering,trajectory frequent mode and abnormality detection.Although there are many clustering algorithms and trajectory anomaly detection algorithms,the main idea is to spatially cluster and anomaly detection of trajectory data,ignoring other information features such as time and geographic information.In order to solve the above problems,the trajectory-based spatio-temporal feature clustering is taken as the starting point,and the trajectory anomalous behavior pattern is found as the goal.An abnormal behavior pattern detection method based on trajectory feature clustering is proposed.The specific work is as follows:(1)A trajectory clustering method based on temporal and spatial features of trajectories is proposed.At present,trajectory clustering usually performs trajectory segmentation to form trajectory segments,and then similarity measurement methods are used to measure trajectory segments.The similarity is high.The method takes a long time and the space is also very high..Aiming at these problems,a trajectory clustering algorithm based on temporal and spatial features of trajectory is proposed.The time-based feature-based coarse clustering algorithm is used for each trajectory in the trajectory.Then the space-time features of each trajectory are calculated to form a vector space.The clustering method is performed by the classical clustering algorithm to obtain the final clustering.(2)An anomaly detection method based on trajectory clustering is proposed.The trajectory cluster obtained by the clustering method provided above can calculate the density value and the time abnormal threshold of the trajectory field.These two can be used for abnormality detection,and can be divided into multiple trajectory anomalies,such as time characteristic anomaly and spatial characteristic anomaly.And global exceptions.Further,since the new sampling data in the trajectory data will continuously arrive in real time,the data in the anomaly detection algorithm is updated,and a sliding window-based algorithm is proposed,and an abnormality detection is performed every time a certain data is collected.Finally,the effectiveness of trajectory clustering and anomaly detection is verified by experiments.By analyzing the experimental results,the method is not only reasonable and effective,but also accurate and effective trajectory clustering and anomaly detection,so as to mine similar A group of moving objects with social roles,behavioral characteristics,and interests.