Principal Components and Neural Networks Based Linear Regression to Determine Biomedical Equipment Maintenance Cost in the Peruvian Social Security Health System

Abstract

In this study, multivariate linear regression models and principal component analysis, and artificial neural networks (ANN) were designed to predict the monthly cost of biomedical equipment maintenance services in the Peruvian Social Health Insurance (EsSalud). The data employed in the development of these models were obtained from maintenance contracts and their execution, from 2019 to present. The results demonstrate that the multivariable linear regression model acquires adequate metrics; still, such a model has four correlated variables. Hence, the use of the principal component regression model enhanced the outcomes by using two components, thus acquiring greater interpretability. Finally, the ANN model obtained the best performance predictor.