Optimizing Aquilaria species classification for agarwood production through data analysis and KNN modelling

Aquilaria, a valuable plant species renowned for its production of gaharu, commonly known as agarwood, belongs to the Thymelaeaceae family. Among its 21 species, 17 have been identified to yield agarwood, which holds significant commercial value and finds applications in perfumes, fragrances, incense, and traditional medicine. The current classification of Aquilaria species relies on human knowledge such as evaluation, leading to inconsistent results due to subjective perception and variability in sensory characteristics. For instance, the same agarwood sample may be classified differently by experts based on its aroma, resulting in unreliable species identification. In response, this study proposes an innovative approach for the identification of Aquilaria species using chemical compounds of agarwood oil extracted from GC-FID and classified by KNN classifier. To initiate the process, the identification of significant chemical compounds involves statistical analyses, including boxplot analysis to observe the distribution of Aquilaria species. The compounds present across all Aquilaria species are then selected as significant, employing the z-score test for further validation. Fourteen compounds are identified to exist universally, with three deemed significant through the z-score test. The KNN algorithm serves as the model classifier for Aquilaria species, with all datasets (1st and 2nd datasets) divided into training (80%) and testing (20%) phases. Performance measures, including accuracy, recall, precision, and specificity, are assessed through the confusion matrix. The results demonstrate the KNN algorithm's successful classification of Aquilaria species, achieving 100% accuracy, recall, precision, and specificity in both training and testing phases, meeting performance standard criteria.