Drug Recommender Systems: A Review of State-of-the-Art Algorithms

Abstract

Drug Recommender Systems (DRSs) which are information systems that recommend drug(s) to users based on their symptoms and other factors, have been gaining a lot of research interest recently. These systems help both patients and medical personnel to determine and decide on the best drug prescription with combination to use. Different approaches ranging from machine learning, statistical methods, artificial intelligent, data mining Ontology, matrix factorization etc. have been applied to build a robust DRSs. This paper presents the review of the state-of-the-art algorithms applied to DRS and also gives a summary of a proposed DRS. Findings shows that most recent DRSs use Machine Learning based algorithms such as clustering, sentiment analysis, association rule mining, stacked Artificial Neural Networks, etc., for recommendations. Just a few use other approaches like the Ontology based approach. The DRS reviewed did not take into consideration the feedback from users and most did not consider the peculiarities of patients such as age and pre-existing medical conditions (like allergies and pregnancy) etc, Based on some of the limitations identified, we propose a DRS that will recommend appropriate drugs by considering patients peculiarities. It will also incorporate a feedback mechanism in order to strengthen the knowledge base of the system.