A Data Mining Approach to Identify Sexuality Patterns in a Brazilian University Population.

This paper presents the profile and experience of sexuality generated from a data mining classification task. We used a database about sexuality and gender violence performed on a university population in southern Brazil. The data mining task identified two relationships between the variables, which enabled the distinction of subgroups that better detail the profile and experience of sexuality. The identification of the relationships between the variables define behavioral models and factors of risk that will help define the algorithms being implemented in the data mining classification task.

Mineração de dados aplicada ao conhecimento em uma população universitária / Data Mining applied to knowledge in a university population / La minería de datos aplicada al conocimiento en población universitária

Objetivo: Descrever o processo de mineração de uma base de dados de tabagismo obtida de uma população universitária do Sul de Santa Catarina. Métodos: Estudo de natureza aplicada (tecnológica), transversal, de campo e laboratório, e descritivo. A base de dados utilizada foi de um estudo de prevalência realizado na Universidade do Extremo Sul Catarinense no segundo semestre de 2010, resultando em 575 registros. Foi realizado pré-processamento; em seguida, mineração de dados, primeiro pela clusterização fuzzy, sucedida pela tarefa de classificação; última etapa abordou a avaliação das árvores e regras geradas. Resultados: Foram realizados mais de 300 experimentos, resultando em 524 regras, 339 oriundas da base completa, e 185 da clusterizada de fumantes. Na base completa obteve-se sensibilidade 98,0[IC95%=(97,0;99,0)], especificidade 87,0[IC95%=(97,0;99,0)], acurácia 98,0[IC95%=(79,0;94,0)]; a base clusterizada resultou em sensibilidade 84,0[IC95%=(78,0;90,0)], especificidade 73,0[IC95%=(61,0;86,0)], acurácia 82,0[IC95%=(74,0;89,0)]. Conclusão: O perfil epidemiológico dos tabagistas resultante das regras geradas em nosso estudo foi semelhante da literatura.

Objective: To describe the process of Data Mining of a smoking database obtained from a university population in the South of Santa Catarina. Methodos: Descriptive, laboratory and camp, transversal, technologic nature study. The database used was originated from a prevalence study in the second semester of 2010, at the University do Extremo Sul Catarinense, which has resulted 575 registers. In the beginning the preprocessing was performed; next, the Data Mining, first trough fuzzy clustarization, followed by the classification; last step assessed the generated rules. Results: More than 300 experiments were performed, resulting 524 rules, 339 originated from the complete non-clusterized database, and 185 from smoking cluster. The complete database showed sensitivity 98,0[CI95%=(97,0;99,0)], specificity 87,0[CI95%=(97,0;99,0)] and precision 98,0[CI95%=(79,0;94,0)]; the smoking clusterized database resulted sensitivity 84,0[CI95%=(78,0;90,0)], specificity 73,0[CI95%=(61,0;86,0)] and precision 82,0[CI95%=(74,0;89,0)]. Conclusion: The epidemiologic profile of the tobacco users resultant of the generated rules in our research was similar to the literature.

Objetivo: Describir el proceso de minería de una base de datos del consumo de tabaco obtenido de una población universitaria del Sur de Santa Catarina. Metodos: Estudio de naturaleza aplicada (tecnología), ámbito transversal y de laboratorio, y descriptivo. La base de datos utilizada fue un estudio de prevalencia realizado en Universidad del Extremo Sul Catarinense el segundo semestre de 2010, que resulta en 575 registros. Hemos llevado a cabo pre-procesamiento, a continuación, la minería de datos, primero por agrupamiento difuso, logrado por la tarea de clasificación; etapa final se dirigió a la evaluación de los árboles y las reglas generadas. Resultados: Realizaron más de 300 experimentos, que resulta las normas 524, 339 que surgen de la base completa, y 185 de los fumadores clúster. Base sólida la sensibilidad se obtuvo 98,0[IC95%=(97,0;99,0)], especificidad 87,0[IC95%=(97,0;99,0)], precisión 98,0[IC95%=(79,0;94,0)]; basado en clúster produjo sensibilidad 84,0[IC95%=(78,0;90,0)], la especificidad 73,0[IC95%=(61,0;86,0)], exactitud 82,0[IC95%=(74,0;89,0)]. Conclusión: El perfil epidemiológico de los fumadores que resultan de las normas generadas en nuestro estudio fue la literatura similar.

Using a model of parallel distributed processing associated with data mining in the characterization of sexuality in a university population.

Using the framework for developing parallel applications Java Parallel Programming Framework were conducted performance analysis of an application for the clustering data by the method of fuzzy logic combined with Gustafson-Kessel algorithm. In addition to running in a distributed environment, for comparative purposes, were also conducted collections of processing time in environments with a single Personal Computer approach. With the results obtained by collecting time of application, there was a statistical analysis to validate the application and the algorithm as well as the use of computational clustering as a way to increase performance applications.