Evaluation of Multiple Linear Regression and K-Nearest Neighbor Methods for Predicting Solar Radiation in Saudi Arabia
Abstract: Determining the values of solar radiation is crucial in the design and assessment of solar energy utilisation systems. The aim of this study was to evaluate the effectiveness of the multiple linear regression (MLR) and k-nearest neighbors (k-NN) methods in their ability to predict daily global solar radiation based on meteorological observations. The input parameters were maximum and minimum air relative humidity, maximum and minimum ambient air temperature, and sunshine duration, and solar radiation data were the outputs. The training and testing data sets were formed with data from 10 meteorological stations located in Saudi Arabia. The k-NN method with no distance weighting generated better results than the MLR method. The test dataset correlation coefficients were 0.8170 and 0.6895 for the k-NN method (with the seven nearest neighbours) and for the MLR method, respectively. This study suggests that the k-NN method with k = 7 is suitable for estimating solar radiation where there are no direct measurement devices installed, at least in regions that have similar weather conditions.
Publication year 2020
Pages 433-450
Organization Name
Author(s) from ARC
External authors (outside ARC)
    عبدالرحمن الجنوبي
    عبدالله البخاري
Publication Type Journal
↑