A novel neutrosophic-based machine learning approach for maintenance prioritization in healthcare facilities

Abstract

The development of decision support tools for use in the maintenance management and renewal prioritization of healthcare facility assets is considered a highly challenging task due to the multiplicity of uncertainties and subjectivity levels available in such a decision-making process. Accordingly, this study utilizes a combination of Neutrosophic logic, Analytic Network Process (ANP) and Multi-Attribute Utility Theory (MAUT) to reduce the subjectivity pertaining to expert-driven decisions and produce a reliable ranking of hospital building assets based on their variable criticality levels and performance deficiencies. This is further integrated with the novel use of machine learning algorithms in this field, namely: Decision Trees, K-Nearest Neighbors and Naïve Bayes to automate the priority setting process and make it reproducible diminishing the need for additional expert judgments. The developed model was applied to Canadian healthcare facilities, and its corresponding predictive performance was validated by means of comparison against a previously established model, and its excelling capability was clearly demonstrated. Accordingly, the developed integrated framework is expected to aid in creating a consistent, unbiased and automated prioritization scheme for hospital asset renewals, which in turn is expected to contribute to an efficient, informed and sound resources allocation process.