Minimizing path loss prediction error using k-means clustering and fuzzy logic

Abstract

This research proposes an algorithmic scheme based on k-means clustering and fuzzy logic to minimize path loss prediction error. The proposed k-means fuzzy scheme concurrently utilizes the area topographical variability and multiple path loss prediction models to mitigate the prediction error inherent in the independent use of a conventional path loss model. Vegetation density, manmade structures, and transmission-receiver distances are the fuzzy inputs and the conventional path loss models the output: the free space loss, Walfisch–Ikegami, HATA, ECC-33, Stanford University Interim, and ERICSSON models. The experimental results show that the path loss prediction error of the k-mean fuzzy scheme is only 2.67% compared to the the drive-test measurement, and this is the lowest relative to that of the conventional models. The k-mean fuzzy scheme offers a novel means to approximate path loss in localities with diverse topographical features and also efficiently mitigates the prediction error inherent in the independent use of the conventional prediction models.