<![CDATA[The management of incoming mail archives at a large national logistics company in Indonesia generates a large volume of unstructured textual data, making manual classification inefficient and error-prone. This study evaluates the performance of deep learning models for administrative mail archives classification using data collected between 2023 and 2025. Three models are examined, namely Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Convolutional Neural Network (CNN). Model performance is evaluated using accuracy, precision, recall, F1-score, and confusion matrix metrics. Experimental results indicate that CNN achieves the highest accuracy of 85.82%, outperforming LSTM and Bi-LSTM models. This superior performance is attributed to CNN’s ability to capture local textual patterns through convolution operations, which are well-suited to the structured and repetitive language characteristics of official correspondence. To support practical interpretation, an interactive dashboard is implemented as a visualization tool for model evaluation results, classification outcomes, and clustering analysis. These findings demonstrate that deep learning-based approaches integrated with visual analytics can significantly improve the efficiency and accuracy of unstructured mail archive management]]>
Copyrights © 2026
