A HYBRID GLOVE-BASED FEATURE EXTRACTION APPROACH WITH GAUSSIAN NAIVE BAYES FOR NEWS ARTICLE CLASSIFICATION

This research presents an automated news article classification system that categorizes articles into five distinct categories: business, entertainment, politics, sport, and tech. The system addresses key limitations of manual classification, such as the time-consuming nature of human annotation, subjective inconsistencies, and lack of scalability in handling large volumes of digital content. To overcome these challenges, the proposed method integrates advanced feature extraction techniques with a Gaussian Naive Bayes (GNB) classifier. A user-friendly graphical interface supports the entire classification pipeline enabling users to upload datasets, preprocess them through text cleaning and tokenization, and extract features using a hybrid Global Vector (GloVe) approach. This hybrid method combines global word embeddings with context-aware representations, optimizing classification performance. Experimental evaluation compares three models: the proposed Hybrid GloVe with GNB model achieves the best results with an accuracy of 91.24%, precision of 90.31%, recall of 92.16%, and F1-score of 93.19%. In contrast, the existing PCA with GNB model achieves 79.10% accuracy, 78.71% precision, 78.23% recall, and 78.30% F1-score. The existing LDA with GNB model performs the lowest, with an accuracy of 67.64%, precision of 68.61%, recall of 69.96%, and F1-score of 68.10%. Overall, this automated system significantly improves classification accuracy, reduces processing time, and ensures consistent results making it highly suitable for real-time news classification tasks.