One-Against-All and One-Against-One Multiclass Support Vector Machine Algorithms for Wind Speed Prediction

Abstract

Wind speed prediction has several applications and various atmospheric parameters like temperature, humidity, pressure, wind direction can be used to predict it. A number of methods using mathematical and biological moels have been proposed by various researchers to predict the wind speed. This paper explores the use of one-against-all (OVA) and one-against-one (OVO) multiclass Support Vector Machine (SVM) algorithms for wind speed prediction. The algorithms are tested on wind speed data having hundreds of samples of training and test data sets. The results of employing the two algorithms are compared for predicting the wind speed and results indicate that one-against-one algorithm produces better results than the one-against-all algorithm.