Classification of Fermentation Levels of Cacao Beans (Theobroma cacao L.) Using Sensing Technology and Neural Networks

The global chocolate market is proliferating, and this growth has a significant economic impact on cacao-producing regions. Smallholder farmers in these regions rely heavily on the chocolate industry's income, providing them a much-needed boost. The quality assessment of cocoa beans is essential in the cocoa industry as it directly affects the value and price of the final product. The combination of artificial neural networks with the electronic nose device has emerged as a promising method to assess the quality of cocoa beans. The study classifies the fermentation degree of cacao (Theobroma cacao L.) beans through sensing technology and deep learning. A moving average is applied to the raw data to eliminate volatility. A comparison of pre-processing techniques to better understand the data and identify critical patterns and relationships within the dataset. ANN, DNN, and TabNet were employed as data modeling techniques. Deep Neural Network (DNN) shows remarkable results to the other two architectures. Visualization techniques using a dashboard display gas data graphs and fermentation degree classifications throughout fermentation. The dashboard serves as monitoring to reduce the waste of unfermented cacao beans cut during cut-testing.