Classifying News Articles on Nigerian Newspapers Using Machine Learning

With the escalation of news sources, it presents challenge for a reviewer to find the exact newspaper article(s) that would meet ones need(s); hence, the need to classify news articles so that users can efficiently get the information they need with minimal efforts. The classification of news presents many advantages, such as, determining which newspaper to look out for albeit the date of publication, as well as providing the user with personalized recommendations based on the user's previous areas of interest. Also, for research purposes, it can tell the class of news a particular string belongs to, in addition to finding trending news articles. Studies on classification of news for other countries’ Newspapers have been ongoing, into a variety of groups/categories. Examples are: Arabic, Tigrigna, Uzbek, China, and Thai news headlines have all been classified into a variety of categories. These would help visitors to quickly find the news stories that interest them, little to no research has been done to categorize Nigerian news articles. Therefore, this research proposed the classification of Nigerian news articles into five different categories; such as, sports, entertainment, business, health, and politics using machine learning - to the extent of displaying the class of a string keyed in by the user - as a means of filling-in the gap in the existing body of literature. A local source was used to compile a dataset consisting of five different Nigerian news articles. Feature extraction employed Linear Discriminant Analysis (LDA), and Support Vector Machines (SVM). These two algorithms were used to train the model. The research will also evaluate the outcomes of several categorization algorithms and compare the efficacy of these algorithms according to some of performance metrics including accuracy, precision, and recall.