Determination of support vector regression parameters using African buffalo optimization algorithm

Inusa Sani Maijama’a (University Utara Malaysia)
Yuhanis Yusof (University Utara Malaysia)
Mohamad Farhan Mohsin (University Utara Malaysia)

Indonesian Journal of Electrical Engineering and Computer Science
Vol 28, No 2: November 2022

The use of support vector regression (SVR) for regression tasks has been on increase over the past few years. Unfortunately, the practical application of SVR for regression task is limited due to its dependence on proper setting of its hyper-parameters and associated kernel parameter. Therefore, it become imperative to device a reliable and fast mechanism of determining the value of these parameters that could guarantee lowest generalization error. This paper presents SVR parameter optimization approaches using African buffalo optimisation (ABO) algorithm, i.e. SVR-ABO. The SVR parameters are optimized by using African buffalo optimisation algorithm. Results obtained from several experiments performed has shown that the proposed ABO algorithm has the capability of determining SVR hyper-parameters which most of time has to be done through estimation.