ABSTRACT
This project's main objective is to discover which are those comorbidities that could lead to a fatal outcome in a patient diagnosed with COVID-19 in the state of Baja California through a classification algorithm using neural networks. For this, a database obtained on the federal government portal by the General Directorate of Epidemiology with a cutoff date of June 8, 2020 was used. Only the records of the residents of Baja California were kept and only the following data: Sex, Municipality, Date of death, Age, all those variables referring to morbidities, Result (Confirmed cases of COVID-19), ICU (If they needed to enter the intensive care unit);also, from the variable of the date also, from the date variable of death, another variable called "Deceased" was generated to categorize whether the patient died or not. The resulting database was imported into the software where the model of the neural network, data preparation was performed and built the neural network model (multilayer perceptron). The dependent variable "Deceased" was selected, as variables the variables referring to the patient's comorbidities and as a covariate the variable of the scalar type Age. For this model, a random partition of the data was carried out, where 70% of the data was assigned for training and the remaining 30% for tests, obtaining a success rate of 82% and an 18 % error.