Mapeamento do risco da esquistossomose em Minas Gerais usando k-NN e árvore de decisão

Abstract

Of all the parasitic diseases that affect humans, schistosomiasis is one of the most widespread. Considereda serious public health problem, the disease affects thousands of people in Brazil. Since the implementation of schistosomiasis control program in the state of Minas Gerais, stock control and surveillance have been conducted. To contribute to the control and mapping of endemic areas, the aim of this study is to obtain thematic maps showing the risk factor for schistosomiasis mansoni in Minas Gerais. Schistosomiasis is a disease caused by a worm that uses a snail as intermediary host. The worm uses the water to go from the snail to humans. Several variables can contribute for a high risk of a population contracting the disease. In this study, this risk is evaluated from climate, socioeconomic and remote sensing variables, which include MODIS and SRTM data. In this work, two pattern recognition techniques were used to generate two risk maps, with several parameter configurations.The first one is decision trees, for which a total of 19 classifications were generated. The second one technique isthe nearest neighbour classification. For this method, only the number of neighbours varied, and 11 classifications were generated. Results showed a better result for the decistion trees in most part of the tests.