以C4.5建立病患分類模型之探討－－以兩家醫院之糖尿病病患為例

Use C4.5 to Build a Classifier for the Patients--Take the DM Patients of Two Hospitals as an Example

糖尿病病人常因血糖控制不當， 導致腎、肝、眼、四肢末梢病變等併發 症，所以常利用生化數據來對病情加以監控，許多研究證實臨床撿驗數據和 人種有一定關f系， 但比較少研究分析兩鄰近醫院的就診病患是否有族群、地 域之差異。資料探勘（Data Mining ）之技術導入臨床分析已有許多年，較常 用於因果關係的預測， 鮮少用於找出族群差異之重要因子，本次研究以兩鄰 近醫院第二型糖尿病病患門診檢查數據為探勘之分析資料， 嘗試以-C4.5- 決策 樹（Decision Tree ） 方法來分析這類型病患，研究影響併發症的相關因于其在 不同地域、族群之就診病人間是否有差異，藉以幫助做為疾病治療及衛教的 依據。此研究將兩醫院數據，以蔣以仁博士自行撰寫之JAVA 程式“ Index miner' （ Chiang and Lin, 2001 ） 中之C4.5 分析工具加以分析， 分析結果主現兩 醫院第二型糖尿病門診就診病人， 可用影響併發症的相關因子加以明顯分 類，顯示兩醫院這類型病忠來源有其族群、地域之差異，因此將-C4.5- 決策樹 法應用於臨床分析應有其價值。

The diabetes patients often due to the blood sugar controlled improperly, lead to the pathological change of kidneys, liver, eyes, and four limbs, so need to control the biochemical data. Many researches approve ethnic group have relations with lab data, but has not research to analyze the patient's data of two neighboring hospitals, the technology of Data Mining is introduced into clinic already a lot of years, usually used for predicting, seldom used for the application of population classify, This study use the biochemical data of the clinic out patients, try to analyze two hospital patients who suffer from second type DM, attempt to prove that C4.5 can find out the difference among different hospital patient's population, in order to help to treatment and health teaching. The patient's data were analyzed by JAVA C4.5 tool developed by Chiang. The result find that the tool can obviously separate two groups data and verify that there are significant difference between them.