Telfor Journal Vol.18 No.1 (2026)

Comparative Performance of Mean-Shift and Mode-Shift Outlier Detection of Wireless Channel Multipaths

Topic:

Radio Communications

Abstract

Wireless channel models generate multipaths clustered according to delay, angle of departure, and angle of arrival. The multipaths serve as datasets in the performance evaluation of a proposed wireless communication system. These datasets contain outliers in clustering the multipaths. Outlier removal improves clustering performance. This study compares the performance of mean-shift and mode-shift detectors in identifying outliers in wireless channel model datasets. Mode-shift detected more outliers in the IMT-2020 dataset, whereas mean-shift detected more outliers in the other channel-model datasets. The 2·SD (Standard Deviation) threshold yields the most outliers across all channel scenarios in the datasets. k-medoids performs best on outlier-free datasets, while DBSCAN achieves similar performance on the outlier-free IMT-2020 dataset, after pruning with both outlier methods.

Keywords

Channel model, Clustering algorithm, Multipath clustering, Outlier detection

Full Text

 Comparative Performance of Mean-Shift and Mode-Shift Outlier Detection of Wireless Channel Multipaths

Page(s)

2-5