Nepali news classification using Naïve Bayes, Support Vector Machines and Neural Networks
Tej Bahadur Shahi, Ashok Kumar Pant · 2018
Automated news classification is the task of categorizing news into some predefined category based on their content with the confidence learned from the training news dataset. This research evaluates some most widely used machine learning techniques, mainly Naive Bayes, SVM and Neural Networks, for automatic Nepali news classification problem. To experiment the system, a self-created Nepali News Corpus with 20 different categories and total 4964 documents, collected by crawling different online national news portals, is used. TF-IDF based features are extracted from the preprocessed documents to train and test the models. The average empirical results show that the SVM with RBF kernel is outperforming the other three algorithms with the classification accuracy of 74.65%. Then follows the linear SVM with accuracy'74.62%, Multilayer Perceptron Neural Networks with accuracy 72.99% and the Naive Bayes with accuracy 68.31%.