{"title":"DynMDL: A Parallel Trajectory Segmentation Algorithm","authors":"Eleazar Leal, L. Gruenwald","doi":"10.1109/BigDataCongress.2018.00036","DOIUrl":null,"url":null,"abstract":"The purpose of trajectory segmentation algorithms is to replace an input trajectory by a sub-trajectory with fewer points than the input, but that is also a good approximation to the original trajectory. As such, trajectory segmentation is an essential pre-processing step for trajectory mining algorithms, such as clustering. Among the segmentation strategies that are commonly used for trajectory clustering is Minimum Description Length (MDL)-based segmentation, which consists in finding a sub-trajectory such that the sum of its distance to the input trajectory and its overall length is minimum. However, there are no efficient algorithms for optimal MDL-based segmentation; there are only approximate algorithms. In this work we fill this gap by proposing a parallel multicore algorithm for MDL-based trajectory segmentation. We use three real-life datasets to show that our algorithm achieves optimal MDL, and compare its performance against Traclus, the state-of-the-art approximate Description Length (DL) segmentation algorithm.","PeriodicalId":177250,"journal":{"name":"2018 IEEE International Congress on Big Data (BigData Congress)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"6","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 IEEE International Congress on Big Data (BigData Congress)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/BigDataCongress.2018.00036","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 6
Abstract
The purpose of trajectory segmentation algorithms is to replace an input trajectory by a sub-trajectory with fewer points than the input, but that is also a good approximation to the original trajectory. As such, trajectory segmentation is an essential pre-processing step for trajectory mining algorithms, such as clustering. Among the segmentation strategies that are commonly used for trajectory clustering is Minimum Description Length (MDL)-based segmentation, which consists in finding a sub-trajectory such that the sum of its distance to the input trajectory and its overall length is minimum. However, there are no efficient algorithms for optimal MDL-based segmentation; there are only approximate algorithms. In this work we fill this gap by proposing a parallel multicore algorithm for MDL-based trajectory segmentation. We use three real-life datasets to show that our algorithm achieves optimal MDL, and compare its performance against Traclus, the state-of-the-art approximate Description Length (DL) segmentation algorithm.