Research On Trajectory Prediction Algorithm Based On Sequential Pattern Mining

With the development of positioning technology and the popularity of GPS-equipped smart devices,huge-scale trajectory data has been generated in recent years.And with the continuous development and progress of big data processing methods and people's changing market needs,people can gradually discover and dig out the value of these trajectory data.For example,it can also accurately reflect its movement characteristics and laws while revealing the historical trajectory of a moving object.Therefore,more and more researchers at home and abroad have shifted their research emphasis to the problem of moving object trajectory prediction.Aiming at the massive trajectory data of moving objects on the Internet,this study proposes a Prefix PTPA trajectory prediction algorithm.The main idea of the algorithm is to perform data cleaning and preprocessing and propose a trajectory processing algorithm based on latitude and longitude classification.The operation of stopping point detection and clustering is carried out,in which an improved DBSCAN density clustering algorithm is proposed in trajectory clustering.A series of data processing finally transforms the GPS trajectory set into a trajectory sequence,and make predictions based on the sequence pattern mining algorithm.By comparing with classical algorithms,such as Markov prediction models,the experimental results show that Prefix PTPA trajectory prediction algorithm has advantages,and the average prediction accuracy of the algorithm has improved.The following is the main research work of this study:Firstly,there may be cases where the format does not meet the requirements owing to the large size and clutter of the original data set.Based on that,the study proposes a data processing algorithm on the basis of latitude and longitude classification(ABLLC),which removes substandard trajectories and a large number of repeated trajectories from the three dimensions of longitude,latitude and time.Although data preprocessing will waste a certain amount of time,it saves a lot of time for subsequent trajectory prediction research data processing which will generally make the prediction results faster and more accurate.Secondly,a new density clustering algorithm VP-DBSCAN is proposed in the process of stopping point clustering.This algorithm is to improve the shortcomings of the DBSCAN algorithm because the parameters of the algorithm are fixed,so that the algorithm cannot adapt to the uneven density dataset.This study proposes the idea of variable parameters to cluster data sets which is composed of clusters with different densities,making the clustering results more accurate.Finally,this study compares with the trajectory prediction method on the basis of Markov chain,and proposes PrefixPTPA trajectory prediction algorithm based on the sequential pattern mining algorithm Prefix Span.The comparison between experiments and comparison algorithms shows that the average accuracy of the prediction algorithm has been improved to a certain degree,and it can obtain better prediction results.