Improvement of Fuzzy Geographically Weighted Clustering-Ant Colony Optimization Performance using Context-Based Clustering and CUDA Parallel Programming
| Main Authors: | Nurmala, Nila; Statistics Indonesia, Jalan Dr. Sutomo No. 6-8, Jakarta 10710, Purwarianti, Ayu; School of Electrical Engineering and Informatics, Institut Teknologi Bandung, Jalan Ganesa No. 10, Bandung, 40132 |
|---|---|
| Format: | Article info application/pdf eJournal |
| Bahasa: | eng |
| Terbitan: |
ITB Journal Publisher, LPPM ITB
, 2017
|
| Subjects: | |
| Online Access: |
http://journals.itb.ac.id/index.php/jictra/article/view/2936 http://journals.itb.ac.id/index.php/jictra/article/view/2936/2508 http://journals.itb.ac.id/index.php/jictra/article/downloadSuppFile/2936/734 http://journals.itb.ac.id/index.php/jictra/article/downloadSuppFile/2936/766 http://journals.itb.ac.id/index.php/jictra/article/downloadSuppFile/2936/768 |
Daftar Isi:
- Geo-demographic analysis (GDA) is the study of population characteristics by geographical area. Fuzzy Geographically Weighted Clustering (FGWC) is an effective algorithm used in GDA. Improvement of FGWC has been done by integrating a metaheuristic algorithm, Ant Colony Optimization (ACO), as a global optimization tool to increase the clustering accuracy in the initial stage of the FGWC algorithm. However, using ACO in FGWC increases the time to run the algorithm compared to the standard FGWC algorithm. In this paper, context-based clustering and CUDA parallel programming are proposed to improve the performance of the improved algorithm (FGWC-ACO). Context-based clustering is a method that focuses on the grouping of data based on certain conditions, while CUDA parallel programming is a method that uses the graphical processing unit (GPU) as a parallel processing tool. The Indonesian Population Census 2010 was used as the experimental dataset. It was shown that the proposed methods were able to improve the performance of FGWC-ACO without reducing the clustering quality of the original method. The clustering quality was evaluated using the clustering validity index.