Navegando por Palavras-chave "Mineração De Dados"

Agora exibindo 1 - 4 de 4
  • Imagem de Miniatura
    Item
    Acesso aberto (Open Access)
    Análise do conteúdo textual de mensagens provenientes de redes sociais sobre temas de saúde baseado no inter-relacionamento de doenças, medicamentos e sintomas
    (Universidade Federal de São Paulo (UNIFESP), 2020-05-15) Araujo, Gabriela Denise De [UNIFESP]; Pisa, Ivan Torres [UNIFESP]; Universidade Federal de São Paulo
    Background: Analyzing and interpreting data available on the web, whether on social networks, blogs, or editorial sites, establishing relationships, identifying useful and relevant information is a current and significant computational challenge. The information age has favored the availability of a huge amount of data on the web that has naturally become a rich source of information and evidence on various subjects, including health. Objectives: The purpose of this study is to develop a methodological framework to monitor general public health information from social networks and to contribute to the scientific production of health surveillance studies. Methods: The messages containing at least two medical terms were selected using health terms and phrases related to diseases, symptoms, and medications. Data mining techniques, complex networks, and topic modeling were used to analyze health-related discussions on social networks. Results: About 141 million Twitter messages published in the Brazilian territory in 2017 were collected. Around 95 thousand were classified as health-related. Of these, 27% contained terms related to diseases, 56% related to symptoms and 47% to medications. It was possible to explore the relationship between health terms, the strength of connections and their types, and to observe themes that stood out by measuring their relative importance within the network. With the topic modeling technique, popular subjects were identified, and national health campaign events were highlighted. Unexpected topics were also noted; as symptom treatments and food. Conclusion: Users sharing their opinions and experiences on health topics on social networks can assist in monitoring some aspects of public health and collaborate for participatory surveillance, offering a perception to health managers of how people interact with health topics on the web. The results showed that varied topics related to health are discussed in social networks and the methodologies used in this study are efficient to highlight them and make them useful in terms of information.
  • Nenhuma Miniatura disponível
    Item
    Somente Metadadados
    Estudo computacional do refino do óleo de xisto: análise das variáveis do processo
    (Universidade Federal de São Paulo (UNIFESP), 2020-08-21) Marchioli, Willian Alberto Amaro [UNIFESP]; Concha, Viktor Oswaldo Cárdenas [UNIFESP]; Universidade Federal de São Paulo
    In a world of uncertainty and energy dependence, the search for alternative sources to oil is constant. Oil shale reserves are plentiful and available worldwide. Fuel oil can be extracted from shale, which can be used as a substitute for diesel or kerosene, or can be refined to obtain lighter oils and chemical specialties. The main routes for processing shale oil practiced on an industrial scale are distillation and cracking, processes that have operational limitations due to the physical properties of the oil. Within these routes, sensitivity analysis can be applied to identify the main parameters that affect the process output. This work proposes the use of data mining as a tool for sensitivity analysis automatically. The values used in the data mining stage were taken from the process simulation using the Aspen Plus® software. Using automatic learning methods such as regression tree and linear regression implemented on KNIME software, it was possible to obtain interpretable models that present the order of importance of the parameters of a shale refinement plant at the naphtha stream in output of the process. The contribution of each parameter was validated with the regression coefficients and the branch nodes of regression tree, which indicated that pressure and flow of vapor stream in pre-flash are the most sensible parameters, and it was possible to find the configuration that maximazed the naphta output of the process.
  • Imagem de Miniatura
    Item
    Acesso aberto (Open Access)
    Estudo da Trajetória Navegacional do Médico com interesse em Diagnóstico por Imagem em Ambiente Virtual de Aprendizagem no Contexto da Educação Continuada
    (Universidade Federal de São Paulo (UNIFESP), 2020-09-14) Albanez, Maria De Fatima Bazoni [UNIFESP]; Tarcia, Rita Maria Lino [UNIFESP]; Universidade Federal de São Paulo
    This study describes the medical student's navigational trajectory by mining secondary data recorded in the virtual learning environment. The research, on whose data he relies, had an exploratory, descriptive character, with a quantitative approach and involved the understanding of the student's navigation in a set of virtual learning objects, which made it possible to interpret the quantitative data collected through the analysis of the data log records generated by the educational technological system. This research was carried out on Cetrus Diagnósticos, Ltda, using the information stored in the database of students enrolled in self-instructional distance learning courses, on the topic Diagnostic Imaging and who used the virtual learning environment AVA to carry out their studies. The study included all students who accessed the virtual learning environment; and those who did not have access were excluded. Data collection was performed in the AVA - Moodle, based on the records that generated access report data, which allowed to view each of the course activities and the number of times that this activity was accessed. The data analysis process took place by mining the selected data from the fields of interest and generating the infographics, using the statistical package SPSS - Statistical Package and the spreadsheet software MS-OFFICE Excell®. With support in the planned records, it was possible to obtain statistics from the student's navigational data that allowed to map the choices that the student made during his studies. By knowing how this process occurred, through the analysis of students' navigational behavior, it was possible to unveil a scenario that contributed to the pedagogical designs of the courses, favoring learning and communication, and, in alignment with the interest and motivation of the student. Student, also compete to guide decisions about investments in the production of teaching materials, identify risks and characteristics that indicate the possibility of dropping out, failing or dropping out and to understand digital learning, which is characteristic of contemporary times.
  • Imagem de Miniatura
    Item
    Acesso aberto (Open Access)
    Mapa de interesses de autores de publicações científicas em Informática em Saúde baseado em Modelagem Autor-Tópico
    (Universidade Federal de São Paulo (UNIFESP), 2020-12-18) Baptista, Roberto Silva [UNIFESP]; Pisa, Ivan Torres [UNIFESP]; Universidade Federal de São Paulo
    Introduction: Health Informatics (HI) is an interdisciplinary research field which scientific publication is growing significantly. There are several definitions of HI and research on the thematic structure of HI. Objective: The objective of this research is to present and empirically validate that an application of author-topic modeling and analysis of co-authorship networks in scientific publications makes it possible to highlight interests in research and interaction structures. Methods: Articles from HI journals indexed in PubMed were collected and divided into 5-year periods between 1991 and 2015. An author-topic model was applied to highlight the authors' research interests. Social network analysis techniques were also applied to analyze collaboration between authors. Results: The PubMed retrieval resulted in 69 journals and 76,250 articles. In autor-topic modeling, between 66% and 84% of the topics obtained were labeled. The application of social network analysis techniques showed in all periods that the authors with greater centrality are probably researchers of high relevance in HI. Conclusion: The author-topic models obtained showed robust results, serving as an alternative to evidence the evolution of the HI area from the point of view of the authors' interests identified by the topics obtained. The analysis of coauthorship networks showed the evolution of the global collaboration structure over the years, as well as a local view of the importance of the authors through centrality metrics.