Development of a predictor software to determine the periodicity of patient monitors preventive maintenance in the intensive care unit of a health care institution

This project seeks to develop a software predictor to determine the amount of annual maintenance that a patient monitor (PM) will have in order to adjust the periodicity of preventive maintenance for such equipment. To achieve this, a database is created with which the predictor system is trained using regression methodologies such as: multinomial logistic regression, linear regression and stepwise regression. Principal component analysis (PCA) and feature selection by neighborhood component analysis for regression (FSRNCA) are implemented as dimensionality reduction methods. For software coding, Matlab and Python programming languages are used, and the stepwise regression methodology is implemented as a predictor model together with the PCA dimensionality reduction method, obtaining a success rate of approximately 65% with a standard deviation of approximately ±24.