Montoya-Escobar, Diana P.Soto-Álvarez, Claudia2022-01-202022-01-202021-12Soto-Álvarez, C. (2021). Quantifying The Effects of Biomarkers and Comorbidities in Predicting SARS Cov-2 Associated Mortality in Hospitalized Patients in Mexico. Trabajo de obtención de grado, Maestría en Ciencia de Datos. Tlaquepaque, Jalisco: ITESO.https://hdl.handle.net/11117/7697In this retrospective quasi-experimental, cohort study, the biomarkers, demographics, and clinical characteristics of the adult inpatients with laboratory-confirmed COVID-19 from Hospital Regional 110 (Guadalajara, Mexico) were analyzed who were hospitalized over the year 2020, between April 15 (i.e. when the first patient was admitted) to December 31 and had a definite outcome (discharged or dead), to establish the most important variables for the models. In this study, 5 different Classifiers were used: Random Forest, Support Vector Machine, XGBoost, Naïves Bayes, and Symbolic Classifier to classify the outcome of the patients and also to quantify the effect of biomarkers and comorbidities in predicting SARS-CoV-2 positive associated mortality in hospitalized patients. Also, the Symbolic Transofmer was implemented to try to improve the performance of our model. As the dataset includes a big percentage of missing values, we proposed two models, one excluding the missing values and the other including all the missing values. The Random Forest was implemented to obtain the variable importance, and also to the capacity of the model to handle the missing values. The metrics ROC AUC and Accuracy were used to train the models, along with the Bayesian Optimization to tune the hyperparameters and to measure the performance.engCOVID-19Random ForestBiomarkersinfo:eu-repo/semantics/masterThesis