A RECOMMENDER SYSTEM FOR ACADEMIC PUBLICATIONS USING CONTENT-BASED FILTERING TECHNIQUES

ABSTRACT

Publishing is one of the main activities of researchers globally because it gives an opportunity to share the results of their research works. Additionally, the number and quality of academic papers credited to a researcher are one of the criteria commonly used for rating researchers and for determining their promotion. Publications can be grouped based on types, e.g. books, journals, monographs and conference proceedings but our focus here is on journal publication. Journal houses vary in location, editorial board, scope and fields of research. A major challenge faced by researchers, most especially at the early stages of their career, is the choice of journal where to publish. The recent tagging of some journal outlets as predatory compounds this situation. To the best of our knowledge there is no existing recommender system that can guide researchers on where to publish their manuscripts. The aim of this research was to develop a recommender system that will guide researchers on where to publish their articles. The proposed recommender system employed a hybridized content-based filtering method, which combines content-based filtering, Chi-square and multinomial logistic regression techniques. This system generates recommendations from a datasets obtained with the web crawling algorithm, which was used to continuously updating the training set and the learning model. The Chi-Square model was adopted in selecting vector space and multinomial logistic regression was employed as a classifier for predicting similar features. Experimental results prove that the proposed recommender system satisfies the researcher’s desired outcome. The recommender system was tested with samples of manuscripts’ abstracts from researchers and the outputs were similar to the possible predictions from the research fellows. The proposed tool was able to suggest and recommend the most relevant journals and conferences to the researchers based on the input provided (paper title and abstract). The important part of this work is the ability of the system to recommend journals based on their degrees of impact factor.

Keywords: Impact Factors (IF), Web mining, Stemming, Web crawler, TF-IDF and Multinomial Logistic Regression