Volume 14, No 1, 2017
Designing a Predictive Analytics Solution for Evaluating the Scientific Trends in Information Systems Domain
Babak Sohrabi, Iman Raeesi Vanani and Mohsen Baranizade Shineh
Abstract
Full-text articles are mostly provided in electronic formats, but it still remains a challenge to find deeply related works beyond the keywords and abstract contexts. Identification of related scientific articles based on keywords and abstracts help researchers to find desired and informative content, but it requires the utilization of variety of text analytics (mining) methods to address the need. This study investigates information systems (IS) articles to cluster and evaluate the domain of knowledge of recent Information Systems publications using text clustering methods, and then to predict the exact clusters of knowledge of new articles using classification methods. This categorization and predictive learning approach help the scholars and practitioners to find the most relevant articles for their researches and practical endeavors through an automated mechanism. Articles have been collected from the Scopus repository. The dataset has been narrowed to specific areas of recent information 33 http://www.webology.org/2017/v14n1/a154.pdf systems research. Different text analytics methods such as text normalization, natural language processing (NLP) and clustering algorithms have been applied and the results for each cluster are evaluated by extensive analysis of identified clusters based on their terms frequency and key phrases. Afterwards, different classification algorithms are applied to learn the current clustering and to predict the major subject focus of a newly published article based on the abstract approximation to the previously learned domains of information systems knowledge. The prediction approach helps the scholars identifying the usability of many articles for further research.
Pages: 32-52
Keywords: Text analytics; Text clustering; Document classification; Natural language processing (NLP), Information systems