Gelatin Source Detection Using Deep Learning: Emphasis on MLP and RNN Variants

Gelatin is a widely used ingredient in the food and pharmaceutical industries, and its classification based on animal source is essential, particularly for ensuring quality and compliance with regulations such as halal requirements. Although accurate, conventional methods for gelatin classification, such as spectroscopic analysis, are time-consuming and costly. With advancements in technology, deep learning models offer a more efficient approach. This study performs a comparative analysis of five deep learning models—namely, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (RNN), Bidirectional RNN, and Multi-Layer Perceptron (MLP)—for gelatin classification based on sensor data. The results indicate that the MLP model achieves the highest performance, with an accuracy of 98.73%, significantly outperforming the other RNN-based models. The comparative analysis reveals that the MLP’s ability to accurately classify gelatin based on sensor data makes it the most effective model among those tested. In contrast, the other models, including LSTM, GRU, Simple RNN, and Bidirectional RNN, demonstrate lower accuracy rates, indicating that RNN-based architectures are less suited for this specific classification task. These findings highlight the superiority of the MLP model in achieving reliable and precise gelatin classification and underscore its substantial advantages for real-time industrial applications.