Comparação de algoritmos de aprendizado de máquinas para desenvolvimento de um sistema para triagem de adolescentes obesos utilizando variáveis clínicas

Abstract

In recent decades, many countries have entered into a development process in general, due to the globalized world that requires high dynamics, there has been a specific increase in the rate of poor human nutrition, causing a rapid nutritional and epidemiological transition, resulting in in considerable individuals with excess body fat even in adolescence. The high prevalence of excess weight during adolescence has become a major problem in human health in general. Adolescence is characterized as a phase in which the human body develops the most and is directly associated with a range of diseases that can put the health of an overweight individual at risk, so the present study aims to estimate the percentage of body fat in adolescents. In order to achieve this objective, machine learning algorithms were selected that were compared, to Even if you could select the one that performs best and can deliver an overwhelming result in general, the database used to apply the algorithms is obtained from data collected from public school students in the city of São Luís do Maranhão. In 2018, the database contains 772 entries of both genders aged 10 to 19 years. With this data it was possible to evaluate indicators such as: age, gender, body mass, height, waist characteristics, hip characteristics, body characteristics calf and arm functionality, as well as the percentage of body fat acquired through bioimpedance. After applying the algorithms, K-SVM had the best performance, showing potential to be used in a future application