<![CDATA[In the digital era, online news has become one of the primary sources of information, encompassing various categories such as politics, technology, entertainment, and business. The increasing volume of news poses challenges in organizing and categorizing information into relevant categories. This study aims to enhance the accuracy of news text classification through a data augmentation approach based on synonym replacement. The methods employed include text preprocessing for data cleaning, augmentation using synonym replacement to improve data diversity, feature representation using TF-IDF and Word2Vec, and modeling with Neural Networks. Evaluation metrics such as accuracy, precision, recall, and F1-score were used to assess performance. The results indicate that data augmentation can improve model accuracy by up to 95%, with balanced training and validation data distributions. The confusion matrix shows that most data can be correctly classified, although some errors occur in categories with similar features. This study demonstrates that synonym replacement-based data augmentation is effective in improving news text classification performance, particularly for datasets with limited training data.]]>
Copyrights © 2025
