Enhancing logistic regression performance for identifying preservatives in coconut sap using electronic nose

Indonesian coconut sugar has experienced rapid export growth in recent years. However, its primary raw material—coconut sap—is highly susceptible to spontaneous fermentation caused by microbial activity, which degrades its quality. To prevent this, farmers often add natural preservatives such as jackfruit wood powder, mangosteen peel, and calcium oxide (CaO). The uncontrolled use of these additives raises concerns about the quality of coconut sugar products, highlighting the need for a rapid and non-destructive detection method. This study aims to develop a detection model for identifying the presence of preservatives in coconut sap using electronic nose (e-nose) technology combined with machine learning algorithms. A total of 240 samples were collected—120 untreated and 120 treated with preservatives. The e-nose device, equipped with ten MOS gas sensors, recorded the volatile compound profiles. Feature extraction was conducted using the integral method, followed by pattern analysis through Principal Component Analysis (PCA). Six classification models were evaluated: Logistic Regression (LR), Naive Bayes, Linear Discriminant Analysis, k-Nearest Neighbors, Support Vector Machine, and Random Forest. Among these, LR demonstrated the best performance, achieving 98.61% test accuracy, near-perfect precision and recall, and an AUC-ROC value of 1.00. The findings indicate that the optimized LR model provides an accurate and scalable solution for detecting preservatives in coconut sap and supports efforts to maintain the quality of coconut sugar production.