Examinando por Autor "Macana González, Carlos Fernando"
Mostrando 1 - 1 de 1
Resultados por página
Opciones de ordenación
- ÍtemIdentificación y pronóstico de sífilis congénita mediante técnicas de Aprendizaje Automático para las localidades de Usme, Tunjuelito, Ciudad Bolívar y Sumapaz (Bogotá D.C.)(Fundación Universitaria Los Libertadores. Sede Bogotá., ) Macana González, Carlos Fernando; González Veloza, José John FredyCongenital syphilis is a serious bacterial infection transmitted in a newborn from a mother who was not treated or was inadequately treated for syphilis during pregnancy; the consequences of this infection in the baby are related to an affectation in the quality of life and diseases such as abdominal masses, low weight, skeletal abnormalities and bone pain, joint inflammation, blindness, deafness, among others, and even death, so it is a problem of interest in public health worldwide; This has led governments and scientists to search for strategies to reduce new cases of syphilis in infants; hence the importance of having predictive models as a tool for early identification of risk factors or variables in pregnant women and thus perform a health action to prevent the transmission of syphilis to the newborn. From this point of gravity and impact that congenital syphilis generates, the present work used machine learning techniques for the elaboration of predictive models that support the identification of variables related to the appearance of new cases of infected newborns and that are useful in health institutions for the timely management of treatment in pregnant women; this from the knowledge of sociodemographic and health variables of the mother and her context. A data set was available that compiles sociodemographic and health information of a cohort of 451 pregnant women with positive diagnosis for syphilis; basic information was available about the newborn in terms of weight and syphilis infection status; it was identified in the data set that 21.5% (n=97) of the births of mothers with syphilis were also born with syphilis (congenital syphilis); 12 prediction models of congenital syphilis were trained using supervised automatic learning techniques. The main result has been to generate four predictive models, K Neighbors Classifier, Light Gradient Boosting Machine, Gradient Boosting Classifier and Random Forest Classifier. The performance metrics of the predictive models were evaluated to select the best of them, achieving an F1-Score of 77.28% in the model based on K Neighbors Classifier, 73.69% in the model based on Light Gradient Boosting Machine, 73.76% in the model based on Gradient Boosting Classifier and 68.38% in the model based on Random Forest Classifier, also with sensitivity above 70%, exceeding the performance metrics of an initial model based on rules; are considered as relevant variables in the predictive potential of the model based on machine learning algorithms: the number of weeks of gestation at the time of the first prenatal checkup; the age of the mother and the; origin of the mother and the number of expected total prenatal checkups.