Low-Cost Voltammetric Electronic Tongue for Classification of Rawon Seasoning Formulation Using Machine Learning

Consistency in seasoning formulation is important for maintaining product quality and consumer acceptance. This paper presents a low-cost electronic tongue system based on cyclic voltammetry (CV) for the classification of rawon seasoning formulation levels. Rather than directly evaluating perceived flavor quality, the proposed approach distinguishes formulation-related compositional differences through electrochemical fingerprints. The system employed five non-precious metal electrodes, namely titanium, stainless steel, copper, nickel, and brass, to generate multidimensional electrochemical responses. Three formulation levels (Standard, Moderate, and Low) were prepared by varying the seasoning amount relative to the manufacturer's recommended formulation. Each sample was represented by 9,370 features extracted from the full voltammograms, forming a high-dimensional electrochemical dataset. Classification was performed using Support Vector Machine (SVM) and Random Forest (RF) algorithms. The results showed that SVM achieved an accuracy of 84.67%, while RF achieved 95.33% accuracy on the independent external test set, demonstrating the effectiveness of machine learning for formulation-level classification based on electrochemical fingerprints.