Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications
Clustering algorithms are commonly used in the mining of static data. Some examples include data mining for relationships between variables and data segmentation into components. The use of a clustering algorithm for real-time data is much less common. This is due to a variety of factors, including...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Telecom |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2673-4001/4/1/1 |
| _version_ | 1797858345531473920 |
|---|---|
| author | Darveen Vijayan Izzatdin Aziz |
| author_facet | Darveen Vijayan Izzatdin Aziz |
| author_sort | Darveen Vijayan |
| collection | DOAJ |
| description | Clustering algorithms are commonly used in the mining of static data. Some examples include data mining for relationships between variables and data segmentation into components. The use of a clustering algorithm for real-time data is much less common. This is due to a variety of factors, including the algorithm’s high computation cost. In other words, the algorithm may be impractical for real-time or near-real-time implementation. Furthermore, clustering algorithms necessitate the tuning of hyperparameters in order to fit the dataset. In this paper, we approach clustering moving points using our proposed Adaptive Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm, which is an implementation of an adaptive approach to building the minimum spanning tree. We switch between the Boruvka and the Prim algorithms as a means to build the minimum spanning tree, which is one of the most expensive components of the HDBSCAN. The Adaptive HDBSCAN yields an improvement in execution time by 5.31% without depreciating the accuracy of the algorithm. The motivation for this research stems from the desire to cluster moving points on video. Cameras are used to monitor crowds and improve public safety. We can identify potential risks due to overcrowding and movements of groups of people by understanding the movements and flow of crowds. Surveillance equipment combined with deep learning algorithms can assist in addressing this issue by detecting people or objects, and the Adaptive HDBSCAN is used to cluster these items in real time to generate information about the clusters. |
| first_indexed | 2024-04-09T21:11:55Z |
| format | Article |
| id | doaj.art-583329ee1d504d4c84ceaf92315916ea |
| institution | Directory Open Access Journal |
| issn | 2673-4001 |
| language | English |
| last_indexed | 2024-04-09T21:11:55Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Telecom |
| spelling | doaj.art-583329ee1d504d4c84ceaf92315916ea2023-03-28T14:58:01ZengMDPI AGTelecom2673-40012022-12-014111410.3390/telecom4010001Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming ApplicationsDarveen Vijayan0Izzatdin Aziz1Computer and Information Sciences Department, Universiti Teknologi PETRONAS, Persiaran UTP, Seri Iskandar 32610, Perak, MalaysiaCenter for Research in Data Science (CeRDaS), Universiti Teknologi PETRONAS, Persiaran UTP, Seri Iskandar 32610, Perak, MalaysiaClustering algorithms are commonly used in the mining of static data. Some examples include data mining for relationships between variables and data segmentation into components. The use of a clustering algorithm for real-time data is much less common. This is due to a variety of factors, including the algorithm’s high computation cost. In other words, the algorithm may be impractical for real-time or near-real-time implementation. Furthermore, clustering algorithms necessitate the tuning of hyperparameters in order to fit the dataset. In this paper, we approach clustering moving points using our proposed Adaptive Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm, which is an implementation of an adaptive approach to building the minimum spanning tree. We switch between the Boruvka and the Prim algorithms as a means to build the minimum spanning tree, which is one of the most expensive components of the HDBSCAN. The Adaptive HDBSCAN yields an improvement in execution time by 5.31% without depreciating the accuracy of the algorithm. The motivation for this research stems from the desire to cluster moving points on video. Cameras are used to monitor crowds and improve public safety. We can identify potential risks due to overcrowding and movements of groups of people by understanding the movements and flow of crowds. Surveillance equipment combined with deep learning algorithms can assist in addressing this issue by detecting people or objects, and the Adaptive HDBSCAN is used to cluster these items in real time to generate information about the clusters.https://www.mdpi.com/2673-4001/4/1/1moving pointsHDBSCANcrowd clusteringunsupervised learning |
| spellingShingle | Darveen Vijayan Izzatdin Aziz Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications Telecom moving points HDBSCAN crowd clustering unsupervised learning |
| title | Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications |
| title_full | Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications |
| title_fullStr | Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications |
| title_full_unstemmed | Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications |
| title_short | Adaptive Hierarchical Density-Based Spatial Clustering Algorithm for Streaming Applications |
| title_sort | adaptive hierarchical density based spatial clustering algorithm for streaming applications |
| topic | moving points HDBSCAN crowd clustering unsupervised learning |
| url | https://www.mdpi.com/2673-4001/4/1/1 |
| work_keys_str_mv | AT darveenvijayan adaptivehierarchicaldensitybasedspatialclusteringalgorithmforstreamingapplications AT izzatdinaziz adaptivehierarchicaldensitybasedspatialclusteringalgorithmforstreamingapplications |