Enhancing E-Nose Performance for Ginger Extract Aroma Classification through Sample Heating and Machine Learning Integration Bambang Heru Iswanto (a*), Muhammad Rosyid Suseno (a)
a) Department of Physics, Universitas Negeri Jakarta
Jl. Rawamangun Muka, Jakarta 13220, Indonesia
*bhi[at]unj.ac.id
Abstract
Accurate aroma detection plays a vital role in the classification of ginger extracts, particularly for applications requiring rapid and cost-effective quality control. This study explores the impact of sample heating on the performance of a low-cost electronic nose (e-nose) system designed for aroma-based classification of ginger extracts. The system utilizes a sensor array composed of metal oxide gas sensors from the MQ series, integrated with machine learning algorithms including Support Vector Machine (SVM) and Artificial Neural Network (ANN). Ginger extract samples were tested under various thermal preconditioning conditions to assess how temperature influences sensor responsiveness and classification accuracy. The experimental results indicate that moderate heating significantly enhances the release of volatile organic compounds, resulting in stronger sensor responses and improved discrimination between different ginger extract types. Under optimal heating conditions, classification accuracy improved by up to 15% compared to ambient-temperature analysis, with the ANN model delivering the best performance. These findings suggest that incorporating thermal treatment is an effective strategy to improve the sensitivity and reliability of low-cost aroma detection systems. The integration of controlled heating with machine learning techniques offers a promising solution for advancing ginger extract aroma classification in both research and industrial quality assessment contexts.
