Nombre: EDUARDO RIBEIRO VARGAS DUARTE
Fecha de publicación: 08/06/2020
Junta de examinadores:
Nombre |
Papel |
|---|---|
| ANILTON SALLES GARCIA | Advisor * |
| HELDER ROBERTO DE OLIVEIRA ROCHA | Internal Examiner * |
| JAIR ADRIANO LIMA SILVA | Co advisor * |
| MARIA CARMEN VIANA | External Examiner * |
Sumario: This work aims to support the Healths Department of Espírito Santo to understand better the situation of violence from external cause (violence and accidents), which represents the third cause of death in the general population and the first in the population aged from 1 to 49 years in Brazil. To achieve this goal, it was proposed the creation of a Semantic data model WHERE the data of
the violence victims, made available in textual records, are organized with the purpose of enabling the use of tools to generate intelligence related to the problem of violence. From the database, the objective is to develop a set of tools using Data Science techniques in order to identify, after integration of the information, the possible causes and consequences of these cases of violence and, based on intelligent reasoning techniques, propose actions and measures that can prevent the most common cases of violence and / or mitigate its
consequences. A structured database was developed using the records of violence from external causes in the state of Espírito Santo from 2011 to 2018. With this database it was developed dashboards that helped to analyze the cities with the highest occurrence of incidents, how personal characteristics of the victim (color, age, marital status, gender) affects the cases and how the
relationship between victim and aggressor influences the incidents. The machine-learning algorithm had consistent results in predicting the cases of violence that happened several times. The Support Vector Machine (SVM) classifier had the best performance, with precision 0.7827, recall 0.7009 and F1 score 0.7235. With this study, it was possible to have a more consistent knowledge related to the violence from external causes in the state of Espírito Santo. The machine-learning model proved to be a great tool to predict the cases of violence that happened several times.
