尋找良率不佳的關鍵主因並提升良率為製造業注重的目標，不論在半導體業、光電面板業、工具機產業，其資料參數越來越龐大同時在產品研發製造初期生產樣本稀少，使資料形成高維度問題，而Cheng等學者 [1] 開發出適用於高維度問題的關鍵參數搜尋演算法 (KSA) 架構，透過兩階段方法搜尋出影響良率的關鍵參數。
然而當記錄損壞資料的形式為計數型資料時，在KSA架構中的關鍵參數搜尋演算法TPOGA與ALASSO皆屬於連續型資料的線性迴歸的參數挑選方法，因而產生誤差偏誤。生產路徑規劃同樣為製造業一大重點，而當前以良率進行路徑規劃的方法 [6] 在面對高維度問題時會產生檢定力不足等情況。
綜合兩項目標，本研究提出以KSA架構為基礎的最佳關鍵路徑搜尋演算法 (GPSA)，除了以適用計數型資料的負二項 (NB) 回歸精進既有演算法，同時開發兩階段方法：第一階段挑選關鍵站點的group演算法、第二階段合併相似良率表現的fused演算法。透過面板與半導體封測的實際資料驗證，證實負二項參數挑選方法能夠得到更精確的結果，且GPSA的關鍵路徑具有實際參考價值。

Searching the key factors of yield loss and increasing yield are the main goals of the manufacturing industry. Regardless of the industries, semiconductor, TFT-LCD, or machine tool industry, both large amounts of data parameters and scarcity of production samples in research and development phase will cause the high-dimensional problem in data science. Cheng et al. [1] proposed a scheme of high-dimensional key-variable search algorithm (KSA), to search for the key variables of yield loss via a two-phase method.
The form of defect data belongs to counting data; however, the KSA algorithm, which contains TPOGA and ALASSO in the scheme, is the feature selection method for continuous data, which may lead to statistic errors and generate deviations. The production path planning is also a main emphasis of the manufacturing industry, and the ordinary yield-oriented path planning methods [6] may cause the lack of testing power when a high-dimensional problem occurs.
To solve the two problems mentioned above, this work proposes a KSA-based golden path search algorithm (GPSA). GPSA not only uses the negative binomial regression to enhance the current algorithm, but also develops a two-step method: key-stage selection with group-related algorithm is the first step, and combining machines that have similar performance in yield with the fused algorithm is the second step. With the real data validation from TFT-LCD and semiconductor cases, the negative binomial feature selection method can generate more precise results, and the golden path generated by GPSA can serve as a good reference for users.

F.-T. Cheng, Y.-S. Hsieh, J.-W. Zheng, S.-M. Chen, R.-X. Xiao and C.-Y. Lin, "A scheme of high-dimensional key-variable search algorithms for yield improvement", IEEE Robotics and Automation Letters, vol. 2, no. 1, pp. 179-186, January 2017.
C. Lin, Y. Hsieh, F. Cheng, Y. Yang and M. Adnan, "Interaction-Effect Search Algorithm for the KSA Scheme," IEEE Robotics and Automation Letters, vol. 3, no. 4, pp. 2778-2785, Oct. 2018.
鄭景文。2017。適用於良率改善之高維度關鍵參數搜尋演算法架構。碩士論文。台南：成功大學 製造資訊與系統研究所。
Raymond H. Myers, "Classical and Modern Regression with Applications ," Duxbury Thomson Learning, Second Edition, 1990.
Hartmust Stadtler, Chrustoph Kilger, Herbert Meyr, “Supply Chain Management and Advanced Planning," Springer Texts in Business and Economics, Fifth Edition, Ch.5~14, 2014.
C.-H. Lee, D.-H. Lee, Y.-M. Bae, and K.-J. Kim, “Determining golden process routes in semiconductor manufacturing process for yield management,” 2017 IEEE International Conference on Industrial Engineering and Engineering, 2366-2470, Dec. 2017.
Bryan E. Denham, “Categorical Statistics for Communication Research,” ISBN, Chapter 10, October 2016.
Yang, Shengping and Gilbert G. Berdine.,“The Negative Binomial regression,” The Southwest Respiratory and Critical Care Chronicles, Vol. 3 No. 10, 2015.
Park MY, Hastie T., “L1-Regularization Path Algorithm for Generalized Linear Models,” Journal of the Royal Statistical Society B, 69, 659–677, August 2007.
Ngai Hang Chan, Ching-Kang Ing, Yuanbo Li & Chun Yip Yau, “Threshold Estimation via Group Orthogonal Greedy Algorithm,” Journal of Business & Economic Statistics, Volume 35, 2017.
Ing, C.K., Lai, T.L., “A Stepwise Regression Method and Consistent Model Selection for High-Dimensional Sparse Linear Models,” Statistica Sinica, 21, 1473–1513, 2011.
Yuan, Lin, “Model selection and estimation in regression with grouped variables,” J. R. Statist. Soc. B, 68, Part 1, pp. 49–67, 2006.
Sergey Bakin, “Adaptive Regression and Model Selection in Data Mining Problems,” PhD Thesis . Australian National University, Canberra, 1999.
L. Meier, S. van de Geer, and P. B¨uhlmann, “The group lasso for logistic regression,” Technical Report 131, Eidgen¨oossische Technische Hochschule (ETH), Z¨urich, Switzerland, 2006.
F. R. Bach, R. Thibaux, and M. I. Jordan, “Computing regularization paths for learning multiple kernels,” Advances in Neural Information Processing Systems 17, 2004b.
R. Tibshirani, Michael Saunders, Saharon Rosset, Ji Zhu, and Keith Knight. " Sparsity and smoothness via the fused lasso ". Journal of the Royal Statistical Society, 67(1) pp:91–108,2005.
Bo Xin, Yoshinobu Kawahara, Yizhou Wang, and Wen Gao, “Efficient generalized fused lasso and its application to the diagnosis of Alzheimer’s disease,” in Proceedings of the Twenty-Eighth AAAI Conference on Artificial Intelligence, AAAI Press, 2163–2169, 2014.
Jun Liu, Jieping Ye, Lei Yuan, “An efficient algorithm for a class of fused Lasso problems,” Conference: 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Washington, DC, USA, July 25-28, 2010.
Hellen Chang, “WLCSP Production Process Introduction”, ASEKH PE, Feb, 2011.