Prototipo de análisis de información para el sistema de salud colombiano aplicado a la enfermedad renal crónica utilizando técnicas de aprendizaje computacional

Abstract

This work deals with one of the problems presented by the health sector in Colombia, due to the increase in costs, related to the growing demand for services, mainly by patients with high-cost cataloged diseases, one of them being Chronic Kidney Disease (CKD), in addition to the scarce and deficient analysis of data that is applied in the country, where “Data Science” techniques and computational learning algorithms are little used. To improve the failures of the health system, first, a broad knowledge of the Colombian health insurance model was acquired. The data was obtained from an EPS and a statistical analysis was carried out using exploratory information methods, where in the data preparation phase, selection, cleaning and transformation processes were developed, which allowed obtaining quality data for mining and use of computational algorithms, and for this CRISP-DM was used, as a complete and detailed methodology of the mining process, in addition the main computer tool was the Anaconda Python distribution, which is widely used for Data Science and machine learning. The result was the obtaining of 2 statistical models to recommend for its application in relation to the cost of health services for people with CKD. Those models were the Ridge Regression model, in the first instance, and the Forest or Random Tree Regression model.