應用人工智慧技術於預測試管嬰兒治療成功率之研究

The Application of Artificial Intelligence Technology to Predict the Success Rate of In Vitro Fertilization Treatment

蕭昌軒(Chang-Hsuan Hslao)

指導教授 ： 顧瑞祥

國立虎尾科技大學/管理學院/工業工程與管理研究所/碩士(2010年)

摘要

在醫療體系中，不孕症(Infertility)是一個相當特殊的疾患，因為不孕症雖被醫學定義為一項疾病，卻鮮少有身體上的徵狀與疼痛，甚至對於生命的維繫不具有危害性，不會影響過正常生活的活動能力，但對於想傳宗接代的人而言，這卻是一項嚴重的病症，於是求助於試管嬰兒(In Vitro Fertilization,IVF)。目前為止醫學上試管嬰兒的成功率普遍不高，且費用相當昂貴，研究發現，不孕患者在沒有足夠的訊息作考量下，就貿然進入試管嬰兒治療，他們不僅對於試管嬰兒的施行過程與做法不瞭解，且因懷子心切，往往高估了懷孕成功率，飽受療程所苦。本研究運用資料探勘的技術類神經網路(Neural Network，NN)、支援向量機(Support Vector Machine,SVM)、分類回歸樹(Classification and Regression Tree,CART)結合C5.0決策樹(Decision Tree)於分析試管嬰兒資料庫，產生決策樹、預測IVF成功或失敗的規則和準確度，希望分析結果可提供IVF的醫師和每一組患者作為成功機率參考依據，藉此降低病人在IVF上的費用和療程上所受的苦。研究結果，類神經網路結合C5.0決策樹有87.11％的準確度、支援向量機結合C5.0決策樹有85.21％的準確度、CART結合C5.0決策樹有81.61％的準確度。

關鍵字

資料探勘 ； 試管嬰兒 ； 支援向量機 ； 類神經網路 ； CART` ； C5.0決策樹

並列摘要

Infertility is viewed, in the health care system, as a special condition. There is rarely any pain or physical manifestations associated with infertility. Infertility is not life threatening, nor does it affect anyone’s daily activities. Nevertheless, for couples hoping to have children, infertility is a serious concern and they often seek help through in vitro fertilization (IVF). The cost of IVF has been quite high, but the success rate has not been. Studies have shown that infertile patients who rushed into IVF treatment without sufficient information did not fully understand the processes involved, and due to the eagerness to have children, often overestimated the rate of success. As a result, many patients have suffered enormously from IVF both financially and psychologically. This research applies data mining of Neural Network, Support Vector Machine, Classification and Regression Tree combined C5.0 Decision Tree to analyze IVF database, which can produce decision tree and anticipate the regulation and accuracy of success or failure from IVF, hope that the result can offer as the reference of success rate to doctors and each patients of IVF, it hereby can reduce the expenditure and suffering on IVF treatment. The result shows that the accuracy on the combination of Neural Network and C5.0 Decision Tree is 87.11%, on Support Vector Machine and C5.0 Decision Tree is 85.21%, and on CART and C5.0 Decision Tree is 81.61%.

並列關鍵字

Data Mining ； In Vitro Fertilization ； Support Vector Machine ； Neural Network ； Classification and Regression Tree ； C5.0 Decision Tree

參考文獻

10.吳充平(2005)，應用資料探勘技術於台灣地區國人健康狀況之研究，南台科技大學資訊管理研究所，碩士論文。

5.行政院衛生署國民健康局(2009)，民國九十六年台灣地區-人工生殖施行結果分析報告。

15.張秀玉、郭碧照(2000)，初次與重覆接受試管嬰兒治療不孕症夫妻之心理社會反應，護理研究，8(2)，190 – 201。

11.李從業、張昇平、陳嘉琦(1997)，不孕夫妻的困擾程度、壓力感受及因應策略的比較，護理研究，5(5)，425–438。

4.石建佳(2009)，應用資料探勘技術於試管嬰兒成功率之預測與輔助診斷，國立虎尾科技大學工業工程與管理研究所，碩士論文。