Abstract—The motivation behind this study is to enhance news text classification by comparing deep neural network models. The problem addressed is the challenge of identifying optimal combinations of text processing methods and vector representations to achieve high accuracy, precision, and F1-Scores. The study aims to identify optimal algorithms and text vector representation combinations to achieve the highest accuracy, precision, and F1-Score performance. The approach involves using the 20NEWS dataset and applying a Bidirectional Long Short-Term Memory-Convolutional Neural Network (BL-CNN) architecture integrated with various pre-trained word embeddings such as Word2Vec, FastText, and GloVe. The results indicate that the BL-CNN model with Word2Vec achieves an accuracy of 92.21%, surpassing other combinations. Notably, the Bidirectional Long Short-Term Memory-Convolutional Neural Network-Bidirectional Long Short-Term Memory (BL-CNN-BL) model with FastText achieves an impressive accuracy of 99.32%, highlighting its superiority in text-processing tasks. The findings underscore the effectiveness of specific vector representations in improving classification performance. In conclusion, the proposed models, particularly the BL-CNN-BL with FastText, demonstrate significant potential for accurate and efficient news text classification, with applications in detecting fake news.
 
Keywords—Bidirectional Long Short-Term Memory (Bi-LSTM), Bidirectional Long Short-Term Memory-Convolutional Neural Network-Bidirectional Long Short-Term Memory (BL-CNN-BL), Convolutional Neural Network (CNN), machine learning, news classification


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).