体育・スポーツ科学分野への決定木分析の応用事例

: 分析方法の紹介と分析の注意点

抄録

　The purpose of this article was to introduce analytic procedures of decision tree analysis in the field of health and sports science. This article dealt with rule search of past sports experiences for establishing exercise habits as an example of decision tree analysis. There are some algorithms in decision tree analysis. A classification and regression tree (C&RT) was selected as an analytic algorithm. The samples were 200 (169 males and 31 females) Japanese university students. Survey items included current exercise habits (frequency, and duration), and past sports experiences (number of sports events, time spent in exercise per week, extent of perceived exercises, and enjoyment of exercise). From the result of a C&RT analysis, the rule of the highest ratio of exercise habit group (75.0%) was enjoyment of exercise > 3, extent of perceived exercises > 4, and number of sports events > 2.5 events (misclassification cost: 34%). It has been reported that a result of decision tree analysis varied depending on analytic algorithms and split criteria. Researchers also reported that samples more than 1,000 needed to show meaningful results. Therefore, the reliability of this result should be examined using different algorithms and split criteria, or different multivariate analysis (e.g., logistic regression analysis). However, it is important for the development of health and sports science to show interim criteria (association rule), although there are some limitations related in analytic procedures.