Estudo e desenvolvimento de métodos para predição de doadores de sangue
Carregando...
Data
2018-02-16
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
Hemotherapy units has difficulties to optimize the search for blood donors in emergency situations,
as well as to keep their blood stocks at adequate levels. On the other hand, the use
of computational techniques for prediction has obtained promissing results in several areas
of knowledge, and can be seen as a fundamental tool in obtaining blood donations, however,
are little explored in this sector. Given this gap, this research aimed to analyze and develop
prediction techniques to optimize the search for donors with higher conversion rate to the
donation, focusing on data mining techniques. For this, we first analyzed the performance
of traditional literature classifiers applied to a real database, which produced unsatisfactory
prediction results. Seeking for higher quality results we propose a top-k recommendation
approach of blood donors, which uses heuristics to estimate a confidence degree in donation.
Computational experiments show that the top-k recommendation approach achieves
good results for all three developed heuristics. The support vector-based heuristic achieving
94.09% of precision among the top-10 recommended, and 99.90% of precision for top-1, for
the same data set that the classifiers were not successful. It is expected that the results of this
research will contribute to the academic community due to the variety of classifiers analyzed
and especially due to the proposed top-k recommendations approach. In the future, this
approach can be better analyzed with other databases and even improved by the development
of new heuristics. In addition, it is believed that the developed top-k approach can
be used in health prediction systems, with a focus on predicting blood donors, especially in
emergency situations.
Descrição
Palavras-chave
Classificadores, Doação de sangue, Sistemas de recomendação, Blood donation, Classifiers, Recommendation systems
Citação
SILVA, F. H. Estudo e desenvolvimento de métodos para predição de doadores de sangue. 2018. 80 f.
Dissertação (Mestrado em Modelagem e Otimização) - Universidade Federal de Goiás, Catalão, 2018.