Multilayer Perceptron (MLP) tuned Kernel Parameter in Support Vector Machine (SVM) for Agarwood Oil Compounds Quality Classification

Mohammad Nur Faisal bin Ariffin*, Nurlaila Ismail, Nor Azah Mohd Ali, Mohd Hezri Fazalul Rahiman, Saiful Nizam Tajuddin and Mohd Nasir Taib

Abstract

Agarwood or gaharu is one of the expensive woods in the earth. The wood is valuable in many cultures due to its extraordinary fragrance and extensively used as perfume ingredient, medicine and incense. The agarwood oil is highly demanded especially in the countries of Saudi Arabia, UAE, China and Japan. As one of the researches in classified the quality of agarwood oil, the implementation of Multilayer Perceptron (MLP) tuned kernel parameter in Support Vector Machine (SVM) are presented in this research especially to classify agarwood oil compounds to the different quality. The works involved of the data taken from previous researcher consist of agarwood oil samples from low and high qualities. The input for agarwood oil was the abundances (%) of compounds and the agarwood oil quality was the output which is low or high. The input and output data of agarwood oil were preprocessed by normalizing, randomizing and splitting the data to training and testing dataset. The training dataset were fed to Support Vector Machine (SVM) for network development model. After that, the testing dataset were used to test on model performance. All the analytical works were performed automatically using MATLAB software version R2015a. The result showed that the Support Vector Machine (SVM) model with Multilayer Perceptron (MLP) tuned kernel parameter that passes all the performance measures; confusion matrix, accuracy, sensitivity, specificity and precision. The finding in this research was benefit the future works of agarwood oil research area especially to the oil quality classifications.