如今,人們的生活型態以及飲食習慣大幅改變,因而引發很多疾病,導致健康亮起紅燈。也因為這些疾病,人們開始注重身體的健康維護,來降低各類疾病發生的機率。本研究是以資料探勘分類技術中的決策樹理論,應用於心臟疾病之預測。經分析後可驗證哪些屬性變數會引起心臟疾病。本研究採用中央研究院人文社會科學研究中心之「民國93-97年度國民營養健康狀況變遷調查」計劃之中總共有3670筆樣本資料,刪除1066筆遺漏值,共採用2604筆資料。因此,本研究採用資料探勘Weka軟體中的方法,包含J48、Naïve Bayes、Multi-Layer Perceptron等演算法,分別比較所檢測出來的準確度結果,並且利用混淆矩陣做分類分析的比較。另外,運用統計分析工具來檢測屬性變數重要程度之結果,來探討心臟疾病相關的病因。進一步地,使用模糊理論來建置模糊專家系統,將模糊專家系統作為罹患心臟疾病診斷之用,並加上心臟疾病的危險因子來建立完整的模糊規則知識庫,讓使用者輸入對應的生理指數,找出符合的模糊規則,最後預測出得心臟疾病的風險評估,進而提供使用者一個決策的依據。本研究結論可以輔助民眾做簡單的心臟疾病預測,並亦可提供給醫師做診斷時之分析參考。
People nowadays, with their changed lifestyle and eating habits, are easy to be afflicted with diseases. To stay healthy, they are now trying hard to avoid getting these diseases. This study adopts the theory of Decision Tree in Data Mining to predict heart disease problems. Through analysis, the attribute variables that cause a heart disease can be identified. This study uses a medical examination database with 3670 medical records from a project titled “National Nutrition and Health Survey in Years 93 to 97” and 1066 data cases with missing values are excluded; finally working on the remaining 2604 valid cases. The other methods involved in this study are J48, Naïve Bayes, Multi-Layer Perceptron. Individual comparisons of the accuracies are made and then Confusion Matrix is introduced in comparing the categorical analyses. The importance, in addition, of the attribute variables is judged using a statistical, analytical tool, which helps locate the cause of heart-related problems. After that, Fuzzy Theory is employed in the foundation of a Fuzzy Expert System to diagnose heart problems. This, when analyzing with risk factors of heart problems, can help build up a complete knowledge base of fuzzy rule. When a user enters a corresponding set of physical indexes, a suitable fuzzy rule can come up to help predict the risk of getting heart problems. The result of this study can help ordinary people do an easy prediction of some heart problems, and be a doctor’s reference in diagnosis.