Evaluating Scaled Conjugate Gradient in NARX Neural Networks: A Comparison with Levenberg-Marquardt for Agarwood Oil Grading

Agarwood oil is highly valued both domestically and globally for its applications in perfumes, incense, and traditional medicine. However, accurately classifying its quality remains a challenge due to the absence of standardized evaluation methods. Existing classification approaches lack consistency and robustness, necessitating more reliable predictive models. This study explores a comparative assessment of the Scaled Conjugate Gradient (SCG) and Levenberg-Marquardt (LM) training algorithms within the Nonlinear AutoRegressive with Exogenous Inputs (NARX) model to determine their effectiveness in grading Agarwood oil. The dataset, obtained from the Forest Research Institute Malaysia (FRIM), Selangor Malaysia, and the BioAromatic Research Centre of Excellence (BARCE), Universiti Malaysia Pahang Al-Sultan Abdullah (UMPSA), is used for model training and validation. A progressive approach is implemented by adjusting the number of hidden neurons to enhance model performance. The models are assessed based on key performance indicators, including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), coefficient of determination (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex>), epochs, and accuracy, to identify the most effective training method for classifying Agarwood oil quality. Computational experiments are conducted using MATLAB software version R2020a. The results indicate that the LM algorithm outperforms SCG based on key performance indicators, demonstrating faster convergence on a dataset of 660 samples. These findings contribute to the advancement of automated Agarwood oil quality evaluation, paving the way for more standardized and efficient grading methodologies.