İşbirlikçi filtreleme ile PubMed makale öneri sistemi

PubMed is one of the largest public databases on biological and medical sciences, it contains more than 30 million biomedical articles cited from several resources such as online books, confrences, and journals, the biggest percentage of citations comes from MEDLINE. In additon to the current articles, PubMed is being updated on a daily basis with new articles. Researchers are finding it very hard to cope with exponentially increasing numbers of biomedical literature, for that reason there is a need to design a recommendation system that helps researchers in finding materials that are relevant to them. In this study we proposed a PubMed article recommendation system, PubGate. Our recommendation system is based on a hybrid approach using both content-based and collaborative approach with focus on the latter. For the collaborative filtering approach, we have used Jaccard similarity to compute the similarities between the users according to their liked articles and their keywords of interest, where we recommended articles that have been liked the most by similar users. Collaborative filtering usually suffers from the cold start problem, which is related to new users who have zero history. To overcome this problem, we integrated Elasticsearch engine to recommend articles to users based on their given keywords of interest. This thesis combines both content-based and collaborative approaches to recommend PubMed articles to the users.