Semantic Trajectory Frequent Pattern Extraction Method And Application Based On Spark

Traveling plays an important role in the daily life of modern people.Users 'travel laws can be mined by studying the spatiotemporal features of people 's trajectories,especially for the study of vehicle trajectories,which can obtain the behavior characteristics of vehicles in various specific time periods.Joint analysis with the user's network attribute tags can also infer important information such as user movement rules,interest preferences,lifestyle habits,route preferences,etc.,which are very useful for understanding the user's behavior patterns,providing high-precision intelligent recommendations,and related information research and decision.Aiming at the problems related to frequent pattern mining of moving trajectories,the paper studies semantic processing of moving trajectories,data cleaning,trajectory compression and frequent pattern mining algorithms based on semantic trajectories.The main content of this article are as follows:First,a compression and merge method based on semantic trajectory is proposed.The original trajectory has the characteristics of data redundancy and noise due to the collection environment and other reasons.In view of this feature,this paper compresses and merges the trajectory data to improve the compression ratio as much as possible while ensuring that valuable information is not lost,so as to simplify the trajectory to the greatest extent.Second,a clustering method based on semantic tree is proposed.This paper extracts the stay points based on the semantic trajectory,while this type of semantic label can be encoded by the tree structure.Also a normalized metric is obtained by improving the Lewinstein distance and can be used to carry out clustering work.Later experiments proved that the clustering is feasible.By controlling appropriate parameters,such as the range threshold,reasonable clustering results can be obtained.Third,the RGP frequent pattern mining algorithm is proposed.The algorithm first selects the frequent item set binomial set,and uses the continuous features and frequent features of the trajectory to make reasonable large-scale branching and pruning.This pruning process brings good efficiency for the later data excavation.In the case of large-scale data and dense data,iterative calculations are very expensive to perform deep iteration operations.This algorithm converts frequent iterative calculations of long item sets into diversity set operations,thereby finding frequent sequences and subsets,and using parallel processing.With spark parallel computing framework,the experiment realizes the mining of frequent patterns of moving trajectories.Comparative experiments prove that the algorithm has better mining efficiency and can realize the mining of multi-user geographic relationships based on the research of individual data.