近年來,社會上的飲食習慣快速變化,導致民眾的身體產生了很多的疾病,也是造成各種大腸異常的主因,有大腸異常症狀的人數在社會上已明顯的上升。因此如何提早發現大腸異常的症狀就是一個很重要的議題。 本研究以陳志達(2012)運用中部某健康中心接受健康檢查和大腸鏡檢查的病患資料為基礎,並應用游雅雯(2008)、林義祥(2011)、陳志達(2012)所找出的影響大腸鏡檢查結果的重要健檢項目集,運用支援向量資料描述(SVDD)與實驗設計,將上述重要健檢項目集作為重要因子,探討其與大腸異常之間的關聯性,建構大腸異常預測模型,並分析SVDD方法的分類預測績效。 研究結果顯示,當(1)綜合上述三位學者的重要健檢項目對分類績效有顯著影響;(2)K疊交互驗證對分類績效無顯著影響;(3)將大腸檢查結果屬於正常定義為目標集(正常)以及將大腸檢查結果屬於管狀絨毛腺瘤及鋸齒狀腺瘤定義為離群集(異常)對分類績效有顯著影響。
In recent years, people’s diet habit has been rapidly changing in society. This is one of the main causes of anomaly colon and colon abnormalities are obviously increased in the society. Therefore, detecting colon abnormalities in the early stage is a very important issue. In this study, the health examination data and the examination data of colonoscopy are used to develop the predicting model of colon abnormalities using the historic data of some related literature, such as Chen (2012), Yu (2008), and Lin (2011). This study applied the support vector data description (SVDD) and experiment factorial Design method to develop the predicting model. The factors of the experiment include the different sets of critical items of the health examination, the number of folds of the cross validation, and the different definition of colon abnormalities. The performance of the predicting model is also studied. The results of this research include: (1) all the critical items obtained from the literature are the critical items; (2) the number of folds of the cross validation is not significant for the performance of the predicting model; (3) the best definition of the target (normal) dataset is assigned from the exactly normal class and the outlier (abnormal) dataset is assigned from the classes of tubular villous adenoma and serrated adenoma.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。