Visualization assisted interactive wireless multipath clustering using dimensionality reduction techniques
Designing wireless communication systems requires a knowledge of the propagation environment which is addressed by using channel models. Cluster-based channel models are nowadays used to develop and evaluate wireless networks based on groups of multipath components (MPCs) with similar parameters cal...
Saved in:
| Main Author: | |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
2022
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etdm_ece/13 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | Designing wireless communication systems requires a knowledge of the propagation environment which is addressed by using channel models. Cluster-based channel models are nowadays used to develop and evaluate wireless networks based on groups of multipath components (MPCs) with similar parameters called clusters. Clustering the MPCs has been widely studied using different algorithms to cluster MPC automatically, resulting in different accuracy. This study improves clustering results through visualization with Dimensionality Reduction (DR) algorithmic techniques namely t-SNE and UMAP and a graphical user interface (GUI) that projects the MPCs to interactively refine the cluster membership accuracy. Generated clustering results from the Simultaneous Clustering and Model Selection Matrix Affinity (SCAMSMA) and the COST 2100 Channel Model (C2CM) data serves as ground truth to test the effectiveness of visualizations along with the Jaccard index and Adjusted Rand Index (ARI) for validation. This work achieves a 0.3368 at 10th percentile, a median of 0.4697, and 0.8884 at 90th percentile of Jaccard membership index for all the datasets, which are vis-a-vis improved the SCAMS result. |
|---|