資料挖礦的技術目前已廣泛運用在許多商業應用領域,例如目標顧客行為預測、市場區隔、信用卡風險控管、市場銷售預測等等。隨著資訊工程的進步,演算法不斷的改良創新,再加上電腦硬體設備功能非常強大,使的挖礦的技術可以轉而運用在科學或醫學上的龐大資料分析。本研究便以臍帶血幹細胞資料庫進行研究探勘,分析臍帶血保存的結果,了解從捐者(產婦)篩檢、生產機構收集到實驗室完成處理及保存的流程中,哪些因數會影響臍帶血最後收集成果的品質,以及保存品質好的臍帶血會在哪些變數的狀態下出現。分析的結果可以提供實驗室未來在事先篩選臍帶血樣本上能夠更為謹慎選擇,確保其保存的臍帶血具有較好的成果。 本研究的目的主要是了解血樣收集結果的概況,探討影響收集結果的因素,以及利用資料探勘技術中屬性歸納法(Attribute oriented-induction)歸納出在臍血捐贈者個人變數、收集醫院變數、及實驗室收集變數上結果規則,了解影響收集結果的變數有哪些,以及保存品質被歸類為屬於優良的臍帶血其變數的特徵描述為何,並將結果與相關的醫學文獻作比對,分析探勘的結果是否與臨床上的統計數據是否有差異,並檢討造成其差異性的原因。 Data mining techniques have been widely applied in many domains including prediction of target customer behavior, market segmentation, sales prediction of products. With the help of information engineering and hardware support, mining techniques also used to be a strong analysis tool for science and medical knowledge. Our research was used to explore data in the database of cell collection facilities and cell procession facilities. From the process of donor evaluation, umbilical cord blood(UCB)collection and UCB procession we wanted to discover characteristic rules for qualifying UCB units with methodology. By methodology for Attribute oriented induction, we can find out what characteristic rules would induce qualifying UCB and compare the results with related reference. Furthermore we can analysis what difference exit and examine the reasons.