A neutrosophic recommender system for medical diagnosis based on algebraic neutrosophic measures

© 2017 Elsevier B.V. Medical diagnosis is a procedure for the investigation of a person's symptoms on the basis of disease. This problem has been investigated and applied to personal healthcare systems in medicine. The relevant methods have limitations regarding neutrosophication, deneutrosophication, similarity measures, correlation coefficients, distance measure, and patients’ history. In this paper, we propose a novel neutrosophic recommender system for medical diagnosis based on algebraic neutrosophic measures. Specifically, a single-criterion neutrosophic recommender system (SC-NRS) and a multi-criteria neutrosophic recommender system (MC-NRS) accompanied by algebraic operations such as union, complement and intersection are proposed. Several types of similarity measures based on the algebraic operations and their theoretic properties are investigated. A prediction formula and a new forecast algorithm using the proposed algebraic similarity measures are designed. The proposed method is experimentally validated on some benchmark medical datasets against the relevant ones namely ICSM, DSM, CARE and CFMD. The experiments demonstrate that the proposed method has better Mean Square Error (MSE) than the other algorithms. Besides, there is no large increase in computational time taken by the proposed method and other algorithms. Experiments by various cases of parameters suggest that the MSE values remain almost the same for each dataset when randomly changing the values of parameters in all the medical datasets. Lastly, the strength of all the algorithms is analyzed through ANOVA one-way test and Kruskal-Wallis test. The proposed method has better accuracy than the related algorithms. Experimental results support the advantage and superiority of the proposed method.