데이터 마이닝 기법을 이용한 저체중 청소년의 특성탐색과 건강증진방안에 관한 연구 - 서울시 고등학생을 중심으로 -

Abstract

The purpose of this paper is to describe general characteristics of underweight adolescents and to search for ways to promote the health of underweight adolescents through assessing health related factors by using data mining techniques. The study sampled(n=4352) 1,180 underweight(BMI<18.5) and 3,172 average weight (18.5<=BMI<23) adolescents, 10th grade students in Seoul, 2000, and investigated the differences between two groups. Related variables were input in a decision tree and an association rule of SAS E-Miner. The most predictable model was CART. In frequency, the proportion of underweight adolescents was higher on the south of the Han-river than on the northern side; but in association rule, associated variables with high support rate and confidence rate were females, north of Han-river, and scoliosis. Therefore, approaches for health promotion of underweight adolescents are not only intervention of physical health, but also the education of proper weight perception to prevent low birth weight and underweight adolescents because mother’s education and c h i l d’s low birth weight are related to underweight adolescents. In conclusion, the following sample groups in Seoul are suggested: female adolescents with scoliosis on the north of Han-river in Seoul.