Os desafios em lidar com dados problemáticos: um estudo em ciência de dados sobre a dengue em Brasília/DF

• Autor(a) principal: Nascimento, João Paulo Cândido
• Data de Publicação: 2019
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/26678
• topic: Ciência de Dados; Data Science; Modelos; Models; Regressão Linear; Linear Regression; Probabilidade; Probability; Dengue; Dengue fever; Clima; Weather; CNPQ::CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::ANALISE DE DADOS; CNPQ::CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::PROBABILIDADE E ESTATISTICA APLICADAS; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::LINGUAGENS DE PROGRAMACAO; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::MATEMATICA DA COMPUTACAO::MODELOS ANALITICOS E DE SIMULACAO

The present work consists of a case study of dengue fever in Brasilia, carried out to explore data science tools to gain insights and discuss the difficulties that exist during a data manipulation process. In this way, data were extracted from official sources of public entities in Brazil on cases of dengue, climatic conditions, demographic censuses and on lending of funds destined to fight epidemiology, all regarding the city of Brasília and for the period of January from 2001 to December 2017. After the data extraction, a process of preparation of the data was necessary, due to inconsistencies found with respect to missing data or problems of granularity of time (some data are daily, other monthly, other annual). Afterwards, the variables were submitted to normality tests, and transformations were made in order to approximate those that did not pass the normal distribution tests. Thus, multiple linear regression models separated by season were constructed, which involved the cases of dengue and the climatic variables with their interactions, where, for summer and autumn, the resulting models were significant to explain the variations in dengue cases , whereas for spring and winter the models found are not sufficient to explain such variations. Analyzes were also carried out involving the transfer of funds and the total number of dengue cases, where it was concluded that, at some moments, the transfer of funds from the current year showed a positive correlation with the cases of dengue of the previous year. Finally, the effect of general characteristics of the population (age, sex and area of ​​residence) was evaluated, trying to determine if these characteristics influence in a greater or lesser probability of contracting dengue. Throughout the work, difficulties encountered during the conduction of a data science analysis were discussed.