Application of artificial neural network in discriminating the agarwood oil quality using significant chemical compounds / Mohd Hezri Fazalul Rahiman … [et al.]

Agarwood oil is well known as expensive oil extracted from the resinous of fragrant heartwood. The oil is in high demand especially in the Middle Eastern countries, China and Japan because of its unique odor. As part of on-going research in grading the agarwood oil quality, the application of Artificial Neural Network (ANN) is proposed in this study to analyze agarwood oil quality using its chemical profiles. The work involves of selected agarwood oil from low and high quality, the extraction of chemical compounds using GC-MS and Z-score to identify the significant compounds as input to the network. The ANN programming algorithm was developed and computed automatically via Matlab software version R2010a. Back-propagation training algorithm and sigmoid transfer function were used to optimise the parameters in the training network. The result obtained showed the capacity of ANN in analysing the agarwood oil quality hence beneficial for the further application such as grading and classification for agarwood oil Objective: To apply ANN in discriminating agarwood oil quality using its significant compounds Problem Statement: The agarwood oil has been sold in various qualities based on its physical appearances such as odour and colour. Consumer perception and high fixative properties also give marks in qualifying the agarwood oil thus affect its price in the market. However this conventional technique makes the price of agarwood oil fluctuate. A standard is needed to ensure that the agarwood oil can be qualified according to its chemical properties so that accuracy can be trusted. Therefore, as part of an on-going research in grading the agarwood oil quality, the application of ANN is proposed to analyse agarwood oil quality using its chemical profiles. Methodology: 1. Data preparation - agarwood oil samples obtained from Forest Research Institute Malaysia (FRIM) 2. Chemical compounds extraction - using GC/MS analysis 3. Component Identification - matching them to the mass spectral library (HPCH2205.L; Wiley7/Nist0.5L; NIST0.5a.L) 4. Data Processing - ANN Application ( Data pre-processing using Z-score, ANN design structure/architecture - parameter optimisation, training and testing the algorithm) Result & Discussion: ANN parameter optimisation - final error for learning rate, momentum rate and hidden layer size ANN final design parameter - Nodes in input layer: 7, Nodes in hidden layer size: 2, Output layer size: 1, learning rate: 0.9, Momentum rate: 0.7, Error goal: 0.01, Epochs: 100 ANN prediction: high accuracy for training and testing prediction (refer to the figure in poster) Patent & List of contributions: 1. Patent no. : LY2013001059 2. UiTM Research Grant: 600-RMI/DANA 5/3/RIF (614/2012) 3. 5 Journals, Scopus indexed: 1. Major volatile chemical compounds of agarwood oils from Malaysia based on Z-score technique - in Chemistry of Natural Compounds, Impact factor: 1.049, Scopus indexed 2. Analysis of high quality agarwood oil chemical compounds by means of SPME/ GC-MS and Z-score technique - in Malaysian Journal of Analytical Sciences, Impact factor: 0.035, Scopus indexed 3. Chemometric study of selected agarwood oils by gas chromatography-mass spectrometer - in Journal of Tropical Forest Science, Impact Factor: 0.537, Scopus indexed 4. Differentiating agarwood oil quality using Artificial Neural Network, in Malaysian Journal of Analytical Sciences, Impact factor: 0.035, Scopus indexed 5. A review study of agarwood oil and its quality analysis - in Jurnal Teknologi, Impact Factor: 0.100, Scopus Indexed. 5.