| Summary: | Preventing fish and beef allergies can be achieved by identifying their components using Surface-Enhanced Raman Spectroscopy (SERS) combined with machine learning techniques. Despite its potential, the application of Support Vector Machine (SVM) in recognizing SERS-based fish and beef gelatin remains unexplored. This paper presents a comparative analysis of SVM performance with and without Principal Component Analysis (PCA) in classifying fish and beef gelatin. Three types of SVM kernels and various PCA stopping criteria were examined to identify the optimal technique. The results showed that the PCA-SVM with the Eigen-One-Criterion (EOC) as the stopping criterion achieved the highest accuracy of 96.53% using the linear kernel. Therefore, the most effective approach for classifying fish and beef gelatin obtained from SERS is PCA-SVM with the Eigen-One-Criterion (EOC) as the stopping criterion, coupled with the SVM linear kernel, which demonstrates superior accuracy, specificity, and sensitivity. © 2024 IEEE.
|
|---|