半導體的良率(Yield)是反應整體產業中的關鍵技術與企業獲利高低之綜合性指標，半導體製造業者除了要有重要的市場策略外，必須在產品品質、交期與生產成本有高人一等的表現，才能在殘酷的市場競爭中立於不敗，而這些決策的制訂及問題釐清皆需仰賴一個兼顧效率性與有效性的良率預測模型。過去在良率管理上，許多學者利用WAT資料來預測良率，其效益是不彰的。另外，大部份業者只使用總缺陷數監測良率，並未考慮缺陷種類(Defect Type)的分析，造成一些特別類型的缺陷異常點可能完全被忽略，不同的缺陷種類組合對於良率產生的殺傷力大不相同，當製程技術不斷創新，晶圓尺寸及元件密度隨之變動，加上不同的生產條件，將使得傳統良率分析模式(ex.Poisson yield model)錯估實際的良率。

綜合上述之問題，為了讓良率管理作出最有效益的生產決策，本研究不以複雜的假設變數或需考量缺陷群聚的現象，利用兩組模擬的晶圓金屬製程缺陷及良率資料，以田口方法找出最佳的基因演算輸入參數組合後，將資料透過十摺交互驗證法(10-fold cross-validation)概念進行基因演算模型之建構，藉以找出晶片所屬的缺陷種類(Defect type)與良率間的關聯及特徵規則，最後的實驗結果與資料探勘中的其它分類模型進行比較，經實驗證明，本研究所提出之方法確實能建構出一有效的良率預測模型及有意義之特徵規則，可供作即時法與事後法的應用，如此將能於早期製造階段提供給工程師作為警示的方向及生產決策的依據，進而節省更多的製造測試成本及增加決策運籌的前置應變時間，提昇企業之競爭優勢。

Semiconductor's yield is a comprehensive target which response to the key technologies and the enterprise profit for whole industry, the semiconductor manufacturing corporation are important not only for commercial strategy but also the outstanding performance with product quality, delivery date and production cost, can’t afford to lose in the brutal market competition, but these decision-making and the question defined depend on an efficient and effective yield model. Many scholars used the WAT information on yield prediction with the yield management in the past, but its benefit wasn’t obvious. Besides, the majority of the enterprise uses the entire defects to monitor yield without consideration of the defect type analysis, that causes some special types of the defects, unusual point possibly to be ignored. The different defect type combination makes entirely different lethality at yield, when the manufacturing ability is innovating continuously, the wafer size and the component density along with its changes plus the different producing condition, will cause the traditional yield model (ex. Poisson yield model) misestimate actual yield.

Above all, in order to make the best of benefit production of the yield management, this research does not use the complex supposition variable or consider the phenomenon of the defects cluster either, using two simulate data on wafer metal process and discovering the best Genetic Algorithms input parameter combination by the Taguchi method, and then using 10-fold cross-validation concept to construct the GA model, so as to discover defect type with the yield relation and feature rule. The final experimental result will compare with other classified models, the result revealed that our approach can construct an effective yield model and extract the meaningful feature rule. This approach can apply to online and offline analysis, that will be able to provide the engineer or decision-maker a basis for early manufacturing direction and decision-making, then reducing more manufacturing cost, increasing arrangement time and raising competitive advantage.

中文文獻部分
1. 林瑞山，「類神經網路於預測晶圓測試良率之應用」，國立成功大學工程管理研究所碩士論文，2003年。
2. 林鼎浩，「建構半導體製程資料挖礦架及其實證研究」，國立清華大學工業工程與工程管理研究所碩士論文， 1999年。
3. 尹柏元，「以遺傳演算法與最近鄰居分類法篩選遺傳疾病基因之研究」，中華民國92年全國計算機會議。
4. 吳明衛，「自動化臉部表情分析系統」，國立成功大學資訊工程研究所碩士論文，2003年。
5. 陳永興，「資料探勘應用於IC設計業良率預測之探討」，銘傳大學管理科學研究所碩士論文，2004年。
6. 陳麗伃，「以資料挖礦為基礎之半導體測試性WAT 資料分析診斷」，國立清華大學工業工程與工程管理研究所碩士論文，2004年。
7. 張喬凱，「利用類神經方法建構晶圓缺陷點群聚圖案之辨識系統」，國立交通大學工業工程研究所碩士論文，2005年。
8. 黃大倫，「半導體製程良率提升專家系統之失效模式與效應&晶圓允收測試異常分析」，國立清華大學工程與系統科學研究所碩士論文，2000年。
9. 黃安橦，「應用支向機於晶圓圖分類之研究」，明新科技大學工程管理研究所碩士論文，2004年。
10. 劉巧雯，「利用資料挖礦方法建立半導體晶圓圖之分群分類系統」，國立清華大學工業工程與工程管理研究所碩士論文，2001年。
11. 劉亭宜，「GRNN 在晶圓製造裡良率模式之建構與分析」，元智大學工業工程研究所碩士論文，1999年。
12. 鄭博文、葉育典、許家豐, 應用田口實驗設計法於基因演算法參數設定-以流程式排程為例. 第一屆台灣作業研究學會學術研討會暨2004年科技與管理學術研討會，頁603~610。
13. 簡禎富,林昀萱,鄭仁傑，「建構模糊決策樹及其在有交互作用之半導體資料之資料挖礦以提昇良率之研究」，品質學報，Vol. 15, No.3. pp.193-209, 2008年。
14. 謝穎欣，「應用田口方法於基因演算法輸入參數設計-以求解多模式專案排程下資源撫平為例」，國立中央大學工業管理研究所碩士論文，2005年。
西文文獻部分
15. Berry, M.J.A. & Linoff, G., “Data Mining Techniques: for Marking, Sales, and Customer Support”, NY: John Wiley ＆ Sons Inc, 1997.
16. Bortfeldt, A.,”A genetic algorithm for the two-dimensional strip packing problem with rectangular pieces”, European Journal of Operational Research, Vol.172, No.3, pp.814-837, 2006.
17. Cavory, G., Dupas, R., & Goncalves, G., “A genetic approach to the scheduling of preventive maintenance tasks on a single product manufacturing production line”, International Journal of Production Economics, Vol.74. pp. 135-146. 2001.
18. Cortes, C. & Vapnik, V., “Support vector networks”, Machine Learning, No. 20, pp. 273-297, 1995.
19. Falcoa, I.D., Cioppab, A. D. & Tarantinoa, E.,"Discovering interesting classification rules with genetic programming", Soft Computing, Vol.1, No. 4, pp.257-269, 2002.
20. Fayyad, U.M., Piatesky, S.G. & Smyth, P., “From Data Mining to Knowledge Discovery in Databases", American Association for Artificial Intelligence, pp.37-54.1996.
21. Gardner, M. & Bieker, J. “Solving Tough Semiconductor Manufacturing Problems Using Data Mining”, Advanced Semiconductor Manufacturing Conference and Workshop IEEE/SEMI, pp.46–55, 2000.
22. Goldberg, D. E., “Genetic Algorithms in Search, Optimization and Machine Learning”, Addison-Wesley, Reading MA, 1989.
23. Guvenira, H. A. & Erelb, E.,”Multicriteria inventory classification using a genetic algorithm”, European Journal of Operational Research, Vol.105, Issue 1, pp.29-37, 1998.
24. Hansen, C.K. & Thyregod, P. “Use of Wafer Maps in Integrated Circuit Manufacturing”, Microelectronics and Reliability, Vol.38, pp.1155-1164,1998.
25. Holland, J. H., “Adaptation in Natural and Artificial Systems”, Ann Arbor: The University of Michigan Press, 1975.
26. Hsua, S.C. & Chien, C.F. “ Hybrid data mining approach for pattern extraction from wafer bin map to improve yield in semiconductor manufacturing”, International Journal of Production Economics, Vol.107, pp.88-103,2007.
27. Kaempf, U., “The Binomial Test: A Simple Tool to Identify Process Problem”, IEEE Transactions on Semiconductor Manufacturing, Vol. 8, No. 2, pp 160-165.1995.
28. Kim,E., Kim, W. & Lee,Y., “Combination of Multiple Classifiers for the Customer's Purchase Behavior Prediction”, Decision Support Systems, Vol. 34, No. 2, pp. 167-175, 2003.[10]
29. Maly, W., Heineken, H.T. & Agricola F., “A sample new yield model”, Semiconductor International, pp. 148-150, 1994.
30. Mitchell, T.M., “Machine Learning”, McGRAW-Hill International Editions.
31. Murata, T., Ishibuchi, H., and Tanaka, H., "Multi-objectivegenetic algorithm and its applications to flowshop scheduling". Computers and Industrial Engineerin, pp.957-968, 1996.
32. Murphy, B.T., “Cost-Size Optima of Monolithic Integrated Circuits”, IEEE Proc, Vol. 52, no. 12, pp. 1537-1545, 1964.
33. Peter, V. Z., “Microchip Fabrication”, McGraw-Hill, New York.1996.
34. Rokach,L.,"Mining manufacturing data using genetic algorithm-based feature set decomposition",International Journal of Intelligent Systems Technologies and Applications,Vol.4, No. 1-2,pp.57-78,2008.
35. Seeds, R.B., “Yield and Cost Analysis of Bipolar LSI”. IEEE International Electron Meeting, pp.12, 1967.
36. Stapper, C.H., “Defect Density Distribution for LSI Yield Calculations”. IEEE Trans. Electron Devices, Vol. 20, pp. 655-657, 1973.
37. Sun, J.U., “A Taguchi Approach to Parameter Setting in a Genetic Algorithm for General Job Shop Scheduling Problem. IEMS, Vol. 6, No. 2, pp. 119-124. 2007.
38. Syswerda, G., “Uniform Crossover in Genetic Algorithms,” In Proceedings of the Third International Conference on Genetic Algorithms, pp. 2-9, 1989.
39. Tan,P.N, Steinbach,M. and Kumar,V. , “Introduction to Data Mining”, Addison-Wesley.2004
40. Tong, L.I. & Chao, L.C. “Novel yield model for integrated circuits with clustered defects”, Expert Systems with Applications, Vol.34, pp.2334-2341, 2008.
41. Witten, I. H., & Frank, E.,” Data mining: Practical machine learning tools and techniques second edition”, Morgan Kaufmann, 2005.
42. Wong, A. Y., “A Statistical Parametric and Probe Yield Analysis Methodology”, Proceedings of the IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems, pp 131-139, 1996.