Location Variability Confidence Interval Model Using Statistical Inference for Over-rooftop Paths and Signal Data Classification Method

In this paper, statistical inference is used to predict a practical linear equation for determining signal loss rates over distance in a suburban over-rooftop path under both line-of-sight (LoS) and non-line-of-sight (NLoS) conditions. Unlike existing prediction equations, the proposed model provides an 80% confidence interval with the same slope as a regression equation and an appropriate intercept based on distance. Additionally, the proposed Gaussian mixture model clustering algorithm can classify unlabeled real-time measurements as either LoS or NLoS with high accuracy, significantly improving the convenience of signal measurement.