Trajectory Representation Learning Based on Road Network Partition for Similarity Computation

Jiajia Li, Mingshen Wang, Lei Li, Kexuan Xin, Wen Hua, Xiaofang Zhou

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

In the tasks of location-based services and vehicle trajectory mining, trajectory similarity computation is the fundamental operation and affects both the efficiency and effectiveness of the downstream applications. Existing trajectory representation learning works either use grids to cluster trajectory points or require external information such as road network types, which is not good enough in terms of query accuracy and applicable scenarios. In this paper, we propose a novel partition-based representation learning framework PT2vec for similarity computation by exploiting the underlying road segments without extra information. To reduce the number of words and ensure that two spatially similar trajectories have embeddings closely located in the latent feature space, we partition the network into multiple sub-networks where each is represented by a word. Then we adopt the GRU-based seq2seq model for word embedding, and a loss function is designed based on spatial features and topological constraints to improve the accuracy of representation and speed up model training. Furthermore, a hierarchical tree index PT-Gtree is built to store trajectories for further improving query efficiency based on the proposed pruning strategy. Experiments show that our method is both more accurate and efficient than the state-of-the-art solutions.

Original languageEnglish
Title of host publicationDatabase Systems for Advanced Applications - 28th International Conference, DASFAA 2023, Proceedings
EditorsXin Wang, Maria Luisa Sapino, Wook-Shin Han, Amr El Abbadi, Gill Dobbie, Zhiyong Feng, Yingxiao Shao, Hongzhi Yin
PublisherSpringer Science and Business Media Deutschland GmbH
Pages396-413
Number of pages18
ISBN (Print)9783031306365
DOIs
Publication statusPublished - 2023
Externally publishedYes
Event28th International Conference on Database Systems for Advanced Applications, DASFAA 2023 - Tianjin, China
Duration: 17 Apr 202320 Apr 2023

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13943 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference28th International Conference on Database Systems for Advanced Applications, DASFAA 2023
Country/TerritoryChina
CityTianjin
Period17/04/2320/04/23

Keywords

  • Representation Learning
  • Seq2Seq Model
  • Similarity Query

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Trajectory Representation Learning Based on Road Network Partition for Similarity Computation'. Together they form a unique fingerprint.

Cite this