Intelligent System for Coffee Bean Roast Level Classification Using Electronic Nose and Artificial Neural Network

Abstract

Roasted coffee beans release gaseous compounds, primarily carbon dioxide (CO₂). Coffee comes in various types, including Robusta, Arabica, Excelsa, Tubruk, Latte, and Luwak. However, this study focuses only on Robusta and Arabica coffee. Each roast level of coffee beans has its own distinct aroma, necessitating a fast and accurate method to differentiate them. Therefore, this research aims to classify coffee bean roast levels based on their aroma profiles. The dataset for classifying Robusta and Arabica coffee roast levels was obtained from data collection using a miniature Electronic Nose system. A total of 900 data samples were collected, with 720 samples used for training and 180 samples used for testing. This study employs an Artificial Neural Network (ANN) with an Electronic Nose for classification. The True Positive (TP) results obtained for each coffee roast level are 44 for Light roast, 55 for Medium roast, and 57 for Dark roast. The classification accuracy achieved in determining the roast level of coffee beans is 86%.