Researchers have developed an Artificial Intelligence model capable of classifying agarwood oil quality with remarkable accuracy. Discover how AI, GC-MS analysis, and neural networks may shape the future of agarwood grading.
The Traditional Challenge of Grading Agarwood Oil
For centuries, agarwood oil has been valued based on experience.
Experienced distillers, traders, and collectors evaluate an oil by observing its aroma profile, complexity, persistence, balance, and overall character. While this traditional approach has served the industry for generations, it also has an unavoidable limitation: human judgement is subjective.
Two experienced evaluators may describe the same oil differently. Personal preference, regional grading standards, and years of experience all influence the final assessment.
As the global agarwood market continues to expand, researchers have been searching for methods that can provide more objective and repeatable quality evaluation.
One promising solution comes from Artificial Intelligence.
A recent scientific study demonstrates that machine learning may help classify agarwood oil quality using its chemical composition rather than relying solely on human perception.
Every Agarwood Oil Has Its Own Chemical Fingerprint
Although every natural agarwood oil smells unique, each oil also possesses a unique chemical profile.
Modern laboratories identify these compounds using Gas Chromatography-Mass Spectrometry (GC-MS), a technique capable of separating and identifying hundreds of volatile molecules present in the oil.
However, analysing dozens or even hundreds of compounds manually is extremely difficult.
Researchers therefore asked an important question:
Can Artificial Intelligence recognise patterns in these chemical compounds that humans cannot easily detect?
Building an AI Model for Agarwood Oil
The researchers developed an Artificial Neural Network (ANN), a machine learning model inspired by the way neurons communicate in the human brain.
Instead of evaluating aroma directly, the model analysed the abundance of selected chemical compounds identified through GC-MS analysis.
To improve efficiency, the researchers first applied Principal Component Analysis (PCA), a statistical method that reduces complex datasets by selecting only the most informative variables.
From many detected compounds, the system identified 11 key compounds that carried the greatest information for quality prediction.
These included:
- α-Guaiene
- β-Agarofuran
- ar-Curcumene
- β-Dihydroagarofuran
- γ-Cadinene
- α-Agarofuran
- 10-epi-γ-Eudesmol
- γ-Eudesmol
- Allo Aromadendrene Epoxide
- Valerianol
- Dihydrocollumellarin
These compounds became the input data for the AI model.
Four Quality Grades
Instead of assigning a numerical score, the researchers classified agarwood oil into four quality categories:
- High
- Medium High
- Medium Low
- Low
The neural network learned to associate different chemical patterns with each grade using a dataset divided into training, validation, and testing groups.
Why the Levenberg–Marquardt Algorithm?
Many machine learning models require extensive training before producing useful results.
In this study, the researchers selected the Levenberg–Marquardt (LM) optimisation algorithm because it is known for fast convergence and efficient learning in neural networks.
Using this approach, the model required relatively few training iterations while maintaining excellent predictive performance.
The Results Were Remarkable
The trained model achieved:
- 100% classification accuracy
- 100% sensitivity
- 100% specificity
- 100% precision
Within the study dataset, every test sample was correctly assigned to its quality category, and the correlation between predicted and actual values reached an ideal value of 1.
These findings demonstrate the potential of AI to recognise complex relationships between chemical composition and agarwood oil quality.
Does This Mean AI Can Replace Experienced Distillers?
Not necessarily.
The researchers themselves acknowledged several important limitations.
The study focused primarily on oils from Aquilaria species and relied on a predefined dataset. It did not evaluate sensory characteristics such as aroma complexity, artistic appreciation, or market desirability.
In other words, AI can analyse chemistry with extraordinary consistency, but it cannot yet replace the human appreciation of fragrance.
A master distiller may detect subtle characteristics that chemistry alone cannot fully explain.
Conversely, AI can provide an objective measurement that helps reduce inconsistency between laboratories and markets.
A Future Where Tradition and Technology Work Together
Rather than viewing Artificial Intelligence as a replacement for traditional knowledge, it may be more useful to see it as a complementary tool.
Imagine a future where:
- GC-MS provides the chemical fingerprint.
- AI estimates the probable quality grade.
- Experienced distillers evaluate aroma, balance, and artistic character.
- Buyers receive both objective laboratory data and expert sensory assessment.
Such an approach could increase confidence in international trade while preserving the craftsmanship that makes natural agarwood so special.
Why This Matters
High-quality agarwood oil commands significant value in global markets.
Objective grading systems could:
- Improve consistency between laboratories.
- Reduce disputes in commercial transactions.
- Assist exporters and importers.
- Support quality control.
- Detect unusual or inconsistent samples.
- Improve transparency for collectors and consumers.
However, no algorithm can fully define beauty.
The fragrance of natural agarwood is influenced by species, geography, tree age, resin formation, distillation technique, storage, and maturation. Many of these factors create sensory experiences that extend beyond numerical classification.
Final Thoughts
Artificial Intelligence is rapidly becoming part of the natural products industry, and agarwood is no exception.
This recent research demonstrates that machine learning can successfully identify relationships between chemical composition and agarwood oil quality with exceptional accuracy under controlled experimental conditions.
For producers, traders, and collectors, this represents an exciting development.
Nevertheless, the finest agarwood oils will likely continue to be appreciated through a combination of science and human expertise.
Technology can measure molecules.
Only experienced noses can truly appreciate the story those molecules tell.
References
This article is based on the 2024 research paper "Accurate Agarwood Oil Quality Determination: A Breakthrough With Artificial Neural Networks and the Levenberg–Marquardt Algorithm" together with related scientific literature cited by the authors.