Prediction of Roasted Coffee Bean Quality Based on Sensor Data

The recent increase in global coffee consumption has heightened the importance of producing high-quality coffee. Coffee roasting, a key stage in the processing of coffee beans, significantly impacts the final quality. The quality index of roasted coffee beans can be calculated using data such as temperature, humidity, and other relevant parameters obtained from roasting machines. This information can be used to improve the quality index before the product reaches consumers. In this study, machine learning and deep learning algorithms were employed to predict the quality index of roasted coffee beans using sensor data collected from roasting machines and processed in a laboratory setting. The algorithms used include Support Vector Regression (SVR), Random Forest Regression (RFR), Decision Tree Regression (DTR), XGBoost, and a Multilayer Perceptron (MLP) neural network. The performance of each model was evaluated using Root Mean Square Error (RMSE), R-squared (R^2), Adjusted R-squared, and Mean Absolute Error (MAE). The Multilayer Perceptron (MLP) has shown the best performance with 7.9976 MAE, 7.6354 RMSE, 0.9132 Adjusted R^2, 0.9132 R^2. Keywords: Machine learning, Deep learning, Regression, Coffee Quality.