การศกษาเปรยบเทยบแบบจำลองทางสถตสำหรบการจำแนกประเภทขาวในสภาพขอมลไมสมดล

Automatic news classification is an important task in natural language processing, as it facilitates the categorisation and retrieval of information from large news sources. This research aims to compare the performance of machine learning techniques by examining the effects of feature extraction methods and imbalanced data handling. The dataset used in this study is the HuffPost News Category Dataset. Data preparation includes text cleaning, stopword removal, and the combination of news headlines with short descriptions. Features are generated using n-grams and transformed into numerical representations using Bag-of-Words (BoW) and term frequency–inverse document frequency (TF-IDF). Four algorithms are evaluated, namely Multinomial Naive Bayes, Complement Naive Bayes, logistic regression, and linear support vector classification (LinearSVC), using 5-fold cross-validation. The experimental results show that linearsvc combined with tf-idf achieves the highest performance, with an accuracy of 82.64% and an F1-score of 81.87%, while multinomial naive bayes is more suitable for BoW. In addition, the use of bigrams helps reduce ambiguity and provides richer textual context than unigrams. For imbalanced data handling, SMOTE produces better results than adasyn and undersampling, achieving an accuracy of 81.67% and an F1-score of 81.68%. In conclusion, this research provides empirical evidence that using tf-idf together with linearsvc and smote for imbalanced data is a highly effective approach for news classification. These findings can be applied to other types of text classification systems and serve as guidance for future research.