Big Data Analysis And Application Based On Vehicle Trajectory

With the rapid development of the logistics industry,the freight demand of logistics companies has increased significantly,so some problems in the industry need to be solved :1)In the process of goods transportation,abnormal behavior such as drivers stealing goods or pulling goods privately occurs,how to detect such abnormal behavior has become a major difficulty in the logistics industry;2)Before delivery,it is necessary to plan the path of the delivery location and find the optimal path,which can reduce the cost of logistics companies and improve the efficiency of goods delivery.The GPS equipment equipped with freight trucks collects a large amount of vehicle trajectory data,which includes information such as vehicle location and driving status,providing the possibility to study the above problems.This paper uses the vehicle trajectory data information to study these two problems,the main contents are:(1)A detection method for heavy truck abnormal parking points is proposed.Firstly,the speed threshold is determined by the speed distribution of the trajectory data to identify the parking trajectory points,and the invalid trajectory points are excluded by the road network matching.Subsequently,then the OD points are identified by clustering the ordered trajectory data,and finally the abnormal OD points are identified based on the isolated forest.Using the vehicle trajectory data of logistics companies to conduct experiments,the results show that the method proposed in this paper can effectively identify abnormal OD points.(2)Improved cuckoo algorithm(DPCS)based on dual population strategy.In order to improve the performance of the cuckoo search algorithm(CS),the algorithm is improved.The chaotic mapping is used to change the initial population generation,so that the initial population can be evenly distributed in the solution space;then propose a dual population strategy,that is,the algorithm initially generates two populations,the first population focuses on the exploration ability,and the second population focuses on the development ability.In the first population,a nonlinear decreasing function is used to replace the step size factor with a fixed value,and the idea of self-learning and social learning of the particle swarm algorithm is used to update the population position.The second population is the CS algorithm mechanism;finally,when the algorithm iterates a certain number of times,the two populations exchange information and eliminate the worse solution,so that the algorithm can improve the optimization accuracy of the algorithm while maintaining the diversity of the population.The experiment results verify the effectiveness of the improved algorithm.(3)The DPCS algorithm is used to solve the path planning problem of logistics distribution.The path planning problem is a discrete optimization problem,which needs to be converted into a continuous problem by random key ecoding.The ecoding is first sorted and mapped into a feasible path,and then the access order of distribution points is finally decided by using the adjacency distance table.Finally,the shortest path can be solved according to the problem model.Using the OD point of the logistics company to conduct a comparative experiment,the results show that the total path planned by the DPCS algorithm is the shortest under the same number of iterations,which indicates that the DPCS algorithm has better optimization ability in the path planning problem,and further proves that the improved algorithm is effective in feasibility in solving real problems.