Research article classification with text mining method

• Published in: Concurrency and computation Vol. 35; no. 1
• Main Authors: Gürbüz, Tuğba, Uluyol, Çelebi
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 10.01.2023; Wiley Subscription Services, Inc
• Subjects: Algorithms; Data mining; Datasets; information extraction; Machine learning; supervised learning; Support vector machines; text mining; web mining; Websites
• ISSN: 1532-0626; 1532-0634
• DOI: 10.1002/cpe.7437

Summary In this article, it is aimed to classify research articles according to their subjects by using text mining method. The data set used for the study consists of research articles that are obtained by writing Python code with a state‐of‐the‐art web mining method. The collection of texts includes articles from the “Medicine”, “Social Sciences”, “Basic Sciences” and “Engineering” topics on the DergiPark‐Academic website. The data set consists of articles written in English and Turkish. In the first phase of text mining, text preprocessing and feature selection steps were applied to text data. Then the articles are classified according to their subject areas by employing Naive Bayes (NB), Random Forest (RF), and Support Vector Machine (SVM) algorithms. In the final phase of text mining, the performance of the algorithms is evaluated with respect to the machine learning performance criteria. The results show that the working times of the algorithms were close to each other. The fastest running algorithm is NB, while the slowest running algorithm is SVM. On the other hand, SVM dominates with the best accuracy results among others. Algorithm performances in both Turkish and English languages do not differ remarkably in terms of accuracy and speed.