Evaluation of RBF and MLP in SVM kernel tuned parameters for agarwood oil quality classification

Abstract

Agarwood oil, famously known as costly oil, extracted from the resinous of fragrant heartwood. The oil is getting high demand in the market especially from China, Vietnam, India, Middle East countries, and Japan because of its unique odour. As one of the researches in grading the quality of agarwood oil, the evaluation of kernel tuned parameter using Radial Basics Function (RBF) and Multilayer Perceptron (MLP) are presented in this paper to classify the quality of agarwood oil by using support vector machine (SVM). The work involved of selected agarwood oil sample from high to low quality. The output was agarwood oil quality either low or high and the input was the abundances (%) of agarwood oil compoundS. The input and output data were pre-processed by following works; data processing (normalisation, randomisation and data splitting into two parts in which training and testing dataset (ratio of 80%:20%) and data analysis using SVM modelling. The training dataset was used to train in developing the SVM model and the testing dataset was used to test/validate the developed SVM model. All the analytical works were performed automatically via MATLAB software version R2013a. The result showed that SVM model with RBF tuning is better than SVM model with MLP tuning and passed all the performance measures; accuracy, precision, confusion matrix, sensitivity and specificity. The finding in this study is significant and benefits further work and application for agarwood oil research area especially its classification.