The capabilities of Multiclass Support Vector Machine (MSVM) training algorithms in grading agarwood essential oil

Anis Hazirah ‘Izzati H., Al-Hadi and Siti Mariatul Hazwa, Mohd Huzir and Nurlaila, Ismail and Zakiah, Mohd Yusoff and Saiful Nizam, Tajuddin and Mohd Nasir, Taib (2024) The capabilities of Multiclass Support Vector Machine (MSVM) training algorithms in grading agarwood essential oil. Malaysian Journal of Chemistry, 26 (1). pp. 314-323. ISSN 1511-2292. (Published)

	Pdf The capabilities of Multiclass Support Vector Machine (MSVM).pdf Restricted to Repository staff only Download (595kB) | Request a copy
Preview	Pdf The capabilities of Multiclass Support Vector Machine (MSVM) training algorithms in grading agarwood essential oil_abs.pdf Download (126kB) | Preview

Abstract

Agarwood essential has a high economic value around the globe for the use of perfumery, medicinal remedies, incense, and other products in the market. However, there is still no standard grading method. Different countries grade agarwood essential oil differently. Traditionally, the grading expert classifies agarwood essential oil by using the aspect of odor, texture, resin color and intensity. Standard grading method is important to ensure the stability of agarwood essential oil's market value. This paper purposes to proof the capabilities of multiclass support vector machine (MSVM) training algorithms in grading agarwood essential oil. Multiclass support vector machine has been identified to be a very effective tool for classification. The MSVM was constructed utilizing radial bias function as the kernel function using MATLAB2021b. One versus all strategies have been added to improve the ability of SVM to classify more than four different grades. The holdout was chosen as the partition for the model with 80:20% training and testing data ratio. The data consists of 660 samples for each significant chemical compound. There are eleven significant chemical compounds which consists of 10-epi-δ-eudesmol, α-agarofuran, β-agarofuran, δ-eudesmol, dihydrocollumellarin, valerianol, ar-curcumene, β-dihydro agarofuran, α-guaiene, allo aromadendrene epoxide and δ-cadinene. The agarwood essential was graded into five grades (low, medium low, upper low, medium high, and high) and six grades (low, medium low, upper low, medium high, high, and upper high). The findings of this paper show the confusion matrix for five grades and six grades have no mismatch between actual and predicted data. The model's performance evaluation results were recorded, with all criteria, including accuracy, sensitivity, specificity, and precision, achieved 100%. In conclusion, the model has the capabilities to identify significant agarwood essential oil chemical compounds and separate agarwood essential oil grades into five and six with high accuracy using eleven significant compounds based on the classification evaluated on two different grades of agarwood essential oil.

Item Type:	Article
Additional Information:	Indexed by Scopus
Uncontrolled Keywords:	Agarwood essential oil; Confusion matrix; Grading technique; Multiclass support vector machine
Subjects:	H Social Sciences > HD Industries. Land use. Labor Q Science > Q Science (General) T Technology > T Technology (General)
Faculty/Division:	Centre of Excellence: Bio-Aromatic Research Center of Excellence Faculty of Industrial Sciences And Technology
Depositing User:	Mr Muhamad Firdaus Janih@Jaini
Date Deposited:	21 Jan 2025 08:37
Last Modified:	21 Jan 2025 08:37
URI:	http://umpir.ump.edu.my/id/eprint/43624
Download Statistic:	View Download Statistics

Actions (login required)

View Item