SVM Modelling of Agarwood Oil Quality Grading using Radial Basis Function (RBF) and Sequential Minimal Optimization (SMO) Learning Algorithms

Abstract

Agarwood Oil also known as Gaharu Oil is an expensive oil with extreme demand in the world trading especially in Japan, China and the Middle East countries. Currently, the grading of the agarwood oil quality only can be done by trained human graders by physically checked the color, odour and the resin content. However, this technique is limited due to human body limitation and not too accurate as it is supposed to be. To improve the problem faced by the existing method, the grading technique using Sequential Minimal Optimization (SMO) and Radial Basis Function (RBF) in Support Vector Machine (SVM) was conducted. The works involved of data collection, data pre-processing, SVM model development and testing of the developed SVM model. The finding showed that both RBF and SMO successfully can grade the agarwood oil quality due to their accuracies is above than 80 % and error rate or MSE is close to 0. Thus, the technique presented in this paper proved its capability in overcoming the constraint of human trained grader and benefit as well as contribute to the agarwood oil industry especially its oil quality grading.