Features extraction based on Naive Bayes algorithm and TF-IDF for news classification

• Published in: PloS one Vol. 20; no. 7; p. e0327347
• Main Author: Zhang, Li
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 30.07.2025; Public Library of Science (PLoS)
• Subjects: Accuracy; Algorithms; Bayes Theorem; Biology and Life Sciences; Classification; Classification systems; Computational linguistics; Computer and Information Sciences; Datasets; Deep learning; Documents; Electronic news gathering; Engineering and Technology; Humans; Identification and classification; Inference; Knowledge representation; Language processing; Latency; Machine Learning; Methods; Natural language interfaces; Natural Language Processing; Neural networks; News; Physical Sciences; Real time; Research and Analysis Methods; Semantics; Social Sciences; Statistical analysis; Support vector machines; Text categorization; China
• DOI: 10.1371/journal.pone.0327347
• ISSN: 1932-6203

The rapid proliferation of online news demands robust automated classification systems to enhance information organization and personalized recommendation. Although traditional methods like TF-IDF with Naive Bayes provide foundational solutions, their limitations in capturing semantic nuances and handling real-time demands hinder practical applications. This study proposes a hybrid news classification framework that integrates classical machine learning with modern advances in NLP to address these challenges. Our methodology introduces three key innovations: (1) Domain-Specific Feature Engineering, combining tailored n-grams and entity-aware TF-IDF weighting to amplify discriminative terms; (2) BERT-Guided Feature Selection, leveraging distilled BERT to identify contextually important words and resolve rare-term ambiguities; and (3) Computationally Efficient Deployment, achieving 95.2% of the accuracy of BERT at 1/52.4th of the inference cost. Evaluated on a balanced corpus of Sina News articles in 11 categories, the system demonstrates a test precision of 95.12% (vs. 84.43% for SVM+TF-IDF baseline), with statistically significant improvements confirmed by 5-fold cross-validation( p < 0.01). The critical findings reveal strong performance in distinguishing semantically distinct categories, while exposing challenges in fine-grained differentiation. The efficiency of the framework (2.1 inference latency) and scalability (linear utilization of CPU resources) validate its practicality for real-world deployment. This work bridges the gap between traditional feature engineering and transformer-based models, offering a cost-effective solution for news platforms. Future research will explore hierarchical classification and the adaptation of dynamic topics to further refine semantic boundaries.