Computation of Radio Refractivity Using Machine Learning Techniques

Abstract

In this paper, different programming languages such as R, MATLAB and Python; framework such as Sci-kit learn; and algorithms such as Regression and Artificial Neural Network (ANN) were used to compute radio refractivity in Abuja, Northcentral Nigeria. The upper air data of atmospheric parameters were collected from the Nigerian Meteorological Agency (NiMet), Abuja, Nigeria with a radiosonde. The data collected are pressure, temperature and relative humidity. The algorithm with the best results for radio refractivity computation was selected, though it was observed that all the algorithms utilised for the computation performed well. However, the ANN performed best as predicted and actual values of radio refractivity were closely related and relative errors were lower compared to other programming languages and algorithms. Hence, it was concluded that using ANN along with MATLAB would be the best algorithm and programming language for computing radio refractivity successfully.