Improved partitioning technique for density cube-based spatio-temporal clustering method

Devi Fitrianah, Hisyam Fahmi, Achmad Nizar Hidayanto, Aniati Murni Arymurthy

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)

Abstract

This work proposes a novel partitioning technique on the density-cube-based data model for the Spatio-temporal clustering method. This work further adapts this clustering approach to Spatio-temporal data. We have compared the IMSTAGRID-the proposed algorithm to the ST-DBSCAN, AGRID+, and ST-AGRID algorithms and have found that the IMSTAGRID algorithm improves the data partitioning technique and the interval expansion technique and is able to achieve uniformity in the spatial and temporal dimensional values. Three types of Spatio-temporal data sets have been used in this experiment: a storm data set and two synthetic data sets – synthetic data set 1 and synthetic data set 2. Both the storm data set and synthetic data set 2 were comparable in terms of the scattering of the data points, while synthetic data set 1 contained clustered data. The performance of the IMSTAGRID clustering method was measured via a silhouette analysis, and its results surpassed the other algorithms investigated; the silhouette index for synthetic data set 2 was 0.970, and 0.993 using synthetic data set data set 1. The IMSTAGRID algorithm also outperformed the baseline algorithms (ST-DBSCAN, AGRID+, and ST-AGRID) in labeling accuracy for the storm data set, yielding results of 82.68%, 38.36%, 76.13%, and 78.66%, respectively.

Original languageEnglish
Pages (from-to)8234-8244
Number of pages11
JournalJournal of King Saud University - Computer and Information Sciences
Volume34
Issue number10
DOIs
Publication statusAccepted/In press - 2022

Keywords

  • Clustering
  • Density-cube spatio-temporal clustering
  • Imstagrid
  • Partitioning technique
  • Spatio-temporal clustering

Fingerprint

Dive into the research topics of 'Improved partitioning technique for density cube-based spatio-temporal clustering method'. Together they form a unique fingerprint.

Cite this