Topic classification is a crucial task where knowledge categories exist within hierarchical information systems designed to facilitate knowledge search and discovery. An application of topic classification is article (e.g., journal/conference paper) classification which is very useful for online submission systems. In fact, numerous online journals/magazine submission systems usually receive thousands of article submissions or even more for each month. This leads to a huge amount of time-consumption of editors to process and categorize the submissions aiming to look for and assign appropriate reviewers to the submitted articles. In this study, we propose an approach based on natural language processing techniques and machine learning algorithms (both classic machine learning and deep learning) to automatic classify the topics of articles in an online submission system. We show by promising performance collected from prediction tasks to present that the proposed method is a potential approach for applying to the real system.