Precise Classification of Five Grades Aquilaria Malaccensis Essential Oil: Multiclass Support Vector Machine Utilizing Pattern Graphical Representation on A Two-Dimensional Graph

Abstract

A member of the Thymelaeaceae family, Aquilaria Malaccensis is a well-known tree species recognized for its aromatic resinous wood. In Indonesia and Malaysia, the tree is known by local names such "gaharu" and "karas”. Its resinous wood is highly valued for its distinct scent and is commonly used in cultural, religious, and economic settings. Grading specialists typically use subjective criteria including odour, texture, resin colour, and intensity to categorize agarwood essential oil. Although qualitative assessments offer useful insights into the quality of the essential oil, the absence of established criteria makes it challenging to guarantee consistency and dependability among various grading procedures. The absence of a uniform grading system weakens market stability for agarwood essential oil. This ambiguity may result in market inefficiencies, pricing differences, and disagreements among buyers and traders. Creating a standardized grading system is vital to tackle these problems and maintain the stability the industry. Implementing a standardized grading system in the agarwood essential oil industry can lead to more market openness, higher customer trust, and improved trade relationships for traders. This study aims to demonstrate the effectiveness of multiclass support vector machine (MSVM) strategies in evaluating agarwood essential oil. The multiclass support vector machine is recognized as a highly successful classification tool. The MSVM was built using a Radial Basis Function (RBF) as the kernel function in MATLAB2021b. There are 660 data samples for each chemical elements in the dataset. There are eleven essential chemical elements in the data samples. The agarwood essential oil was classified into a total of five grades. The research presented in this study demonstrates that the actual and predicted data for five grades do not differ in 5x5 confusion matrix, with the pattern graphical representation being dispersed according to its quality classification. The results of the model's performance measurements were documented, and it met all performance requirements with 100% accuracy, sensitivity, specificity, and precision. In conclusion, using eleven chemical elements based on the classification evaluated on five different grades of agarwood essential oil, the model can accurately identify essential chemical elements in agarwood essential oil and separate agarwood essential oil grades into five.