以決策樹分析與模糊邏輯技術建立肝病罹病風險評估系統

A Risk Evaluation System with Suffering Liver Disease Using Decision Tree Analysis and Fuzzy Logic Technique

現代人愈來愈重視身體保健，多數人也會定期至醫院做健康檢查，然而健檢報告上的資料並非一般大眾所能解讀的。本研究是以資料探勘分類技術中的決策樹分析以及模糊邏輯專家系統應用於肝臟疾病的罹病風險評估，分析案例驗證哪些屬性對於肝臟疾病的判讀是重要指標。本研究採用「民國 93-97年度國民營養健康狀況變遷調查」超過4000筆的體檢資料庫，本資料庫於2013年起開放學術研究單位下載使用，經由諮詢三位醫生專家意見，先選擇出與肝病相關性高的24個屬性。本研究採用屬性顯著性排序及決策樹分析驗證後，篩選出10個與肝病具有顯著相關之屬性，再使用這些顯著屬性以建置模糊專家系統，模糊專家系統可作為罹患肝臟疾病的診斷輔助工具，藉由肝臟疾病之關鍵危險因子建立完整的模糊規則知識庫，使用者可以輸入生理指數與生活習慣，找出符合的模糊規則，系統輸出罹患肝臟疾病的風險評估，本研究可以讓一般民眾輸入健檢資料自行做簡單的肝臟疾病風險評估，也可以提供醫療單位做診斷時的輔助工具。

People nowadays pay more attention to personal health. Most of them go to the hospital for regular checkups. The medical reports, however, are not easy to read for public. This expert system in analyzing cases to see which attributes are vital in judging a liver disease. More than 4000 sample cases from the database, the 2004-2008 survey on the change of health condition of Taiwan people, are adopted and analyzed. The database is released to download for academic research purposes since 2013. This work first selects 24 attributes which are related with liver disease in the database by consulting the comments of three physicians. Then we determine 10 significant attributes via sorting process of attributes and classified verification of decision tree analysis. Furthermore, these 10 attributes are used as fuzzy variables for providing the inputs of a fuzzy expert system to help certify a liver disease case. A fuzzy rule knowledge base is set up through the critical attributes so that a user can input their physical indices and living habits to find an interrelated fuzzy rule. Accordingly, the system can output a risk estimate for liver diseases. This study can help both the ordinary people and a medical center to do their simple, individual checks, or a formal diagnosis.