<![CDATA[This paper explores the application of artificial intelligence (AI), specifically artificial neural networks (ANN), to predict the shelf life of roasted coffee beverages. The research focuses on two advanced AI models: a competitive neural network based on self-organizing principles and a dynamic neural network with time-delayed backpropagation. Both models were rigorously tested and evaluated using several statistical benchmarks, including mean squared error (MSE), root mean square error (RMSE), R² (coefficient of determination), and E² (efficiency coefficient). The results reveal that the competitive neural network outperformed the time-delay backpropagation model. The competitive model achieved an exceptionally low mean squared error of 0.000905264 and a root mean square error of 0.030087601. Its R² value was 0.990947362, indicating a nearly perfect fit, and its E² efficiency coefficient was 0.998759409. In contrast, the time-delay backpropagation model produced a higher MSE of 0.005421264, RMSE of 0.07362923, R² of 0.945787365, and E² of 0.999948307. Using the superior competitive model, the predicted shelf life for roasted coffee drinks was 43 days. This estimate closely matches the observed shelf life of 45 days, demonstrating that ANNs, particularly competitive models, can reliably predict product freshness, potentially benefiting both manufacturers and consumers.]]>
Copyrights © 2026
