Research On Clustering Algorithm And Differential Privacy Protection Algorithm Based On Trajectory Data

With the rapid development of wireless communication,sensor devices,and storage technologies,it is very convenient to obtain the user's current location data and save it to the storage device.The collected trajectory data contains rich personal information.We can provide solutions for various trajectory problems by analyzing and utilizing the trajectory data.At the same time,if the trajectory data is excessively mined and analyzed,or the Internet service providers that hold the user's trajectory data are not restricted,it will inevitably lead to the risk of leakage of the user's personal privacy.Based on the trajectory data,this paper studies the clustering algorithm and the differential privacy protection algorithm respectively.The our work mainly consists of three parts.First,in terms of trajectory clustering algorithm,a Density-based K-means trajectory clustering algorithm(DBK-means)is proposed.First,the distance between all pairs of trajectories is calculated based on the similarity of the trajectory;then,the density value information of each trajectory is counted by using the t-nearest neighbor distance and the eps neighborhood;then,the density value is filtered by the density value ranking.For outlier trajectories with a lower value,considering the density information of the trajectory and the distance information between the current trajectory and the existing cluster center trajectory,the probability of each trajectory being selected as the initial center trajectory is calculated,and the K initial centers with the highest probability are selected;finally,the trajectory clustering is completed after multiple rounds of iterations.The algorithm can avoid the adverse effects of outlier trajectories and random selection of initial center trajectories to the greatest extent.The visualization results and experimental results on synthetic and real datasets show that the clustering accuracy and effect are better compared with existing methods.Second,in terms of trajectory differential privacy protection algorithm,a Differential Privacy Trajectory Data Protection Method Based on Stay Point Allocation of Privacy Budget(DPSP)is proposed.The method first uses time,distance,and point restrictions to find out the location points where the user stays for a long time in a small area;then,the historical sampling points of a single user are clustered to obtain the hotspot areas that the user frequents;then the user's stay point is calculated to the hotspot area The distance from the center and the density information of the hotspot area are considered to obtain the importance score of each stay point;finally,the corresponding privacy budget is allocated to each stay point according to the importance score,and the stay points are added noise that satisfies the differential privacy mechanism called staircase mechanism.We can get the results that the proposed method has better data availability with experiments on real datasets.Third,according to the proposed algorithm,the design and implementation of the desktop application are completed.For the trajectory clustering algorithm,the use scheme of Java FX+Scene Builder framework is adopted;Scene Builder components are used to design and layout the interface UI,and Java FX is used to complete the response events of each component.Similarly,for the trajectory differential privacy protection algorithm,the Py Qt5+Qt Designer framework is used;the Qt Designer component is used to manually drag and stretch components,and the Py Qt5 framework is used to complete the response events of each component.