Accurate Agarwood Oil Quality Determination: A Breakthrough With Artificial Neural Networks and the Levenberg- Marquardt Algorithm

Abstract

The agarwood oil quality has been divided into four grades, including low, medium-low, medium-high, and high, and has been thoroughly examined in this manuscript. Recently, there has been a high demand for agarwood oil but the current grading method is based on conventional techniques that rely on visual inspection of various characteristics such as intensity, smell, texture, and weight. However, this method is not standardized, making it difficult to grade agarwood oil accurately. Therefore, the use of artificial neural networks (ANN) in artificial intelligence (AI) was employed to develop a system for identifying agarwood oil quality using the Levenberg-Marquardt (LM) algorithm. Data from 660 samples of chemical compounds extracted from agarwood oil were used to train the ANN. To enhance the accuracy of agarwood oil quality identification with LM performance, the data was split into 70% for validation, 15% for training, and 15% for testing. The results showed that the ANN with the eleven inputs (10-epi- $\gamma $ -eudesmol, $\alpha $ -agarofuran, $\gamma $ -eudesmol, $\beta $ -agarofuran, ar-curcumene, valerianol, $\beta $ -dihydro agarofuran, $\alpha $ -guaiene, allo aromadendrene epoxide and $\Upsilon $ -cadinene) trained by ten hidden neurons of LM algorithm provided the best performance with 100% for accuracy, specificity, sensitivity and precision as well as minimum convergence epoch. The experimental implementation of the model was done using the MATLAB version R2015a platform. This study will help to standardize agarwood oil quality determination using intelligent modeling techniques and serve as a guide for future research in the essential oil industry.

Keywords

• Aquilaria malaccencis
• ANN
• Levenberg-Marquardt
• multi-layer perceptron
• agarwood oil