2020年全球的新冠疫情，以及美中兩大強權的矛盾。激發出電子用品需求，導致晶片供不應求，2022年台灣IC產業的表現優於全球，工研院產科國際所預估在2022年台灣IC產業產值達4.88兆新台幣，其中台積電營收佔約一半受惠於AI、IoT、車用、高效能運算（High Performance Computing, HPC）等創新應用帶動成長。在晶片供不應求的情況下，半導體的前段製程變得愈加複雜，因此機台的生產效率成為管理者關注的重點。由於機台可能老舊或生產條件及效能不佳，導致原有的前置配方無法滿足要求。因此，除了原廠可以在硬體方面進行改進之外，研究中可過每日的事件記錄和機台配方，以及調整機台的調整束電流參數來進行參數優化。這樣的調整可以提升機台產出的品質，同時提高生產效率。

本研究針對VARIAN VIISTA HC機型提出了一個基於機器學習技術的調整束電流參數成功失敗預測模型。所使用的數據集包含了大量的半導體離子植入的歷史數據，並通過對實驗數據的分析和比較，驗證了該模型的準確性和實用性。為了提升模型的性能，研究中使用了多種機器學習算法進行訓練和優化，包括決策樹推估模式、隨機森林、XGBoost和類神經網路（Artificial Neural Network, ANN）。實驗結果顯示，該預測模型在測試集上的準確率高達90%以上，能夠準確預測半導體離子植入的成功率，並通過優化植入參數提高了生產效率和產品品質。在研究中，研究者提出了一種新的方法，通過限縮Tune Beam Parameters(調整束電流參數)的範圍，有效避免機台可能故障的區間，從而減少機台錯誤的發生。實驗結果證實了這種參數限縮在準確率提高98%以上時對生產效率的有效性。這項研究為半導體製造業提供了一種有效的生產優化方法，同時也為機器學習在半導體工業中的應用提供了一個實例。透過提升企業競爭力和探索新商機，該方法有助於實現企業的長期經營目標。
關鍵詞：資料探勘、預測模型、隨機森林、機器學習、離子植入

The global COVID-19 pandemic in 2020, as well as the tensions between the two major powers, the United States and China, have sparked a demand for electronic products, leading to a shortage of chips. In 2022, Taiwan's IC industry outperformed the global market, with the Industrial Technology Research Institute estimating that the output value of Taiwan's IC industry reached 4.88 trillion New Taiwan Dollars. TSMC accounted for approximately half of this revenue, benefiting from innovative applications such as AI, IoT, automotive, and High-Performance Computing (HPC) that drove growth. In the situation of chip shortage, the front-end processes of semiconductors have become increasingly complex, making production efficiency of the machines a focal point for managers. Due to outdated machines or poor production conditions and performance, the original recipes may not meet the requirements. Therefore, in addition to hardware improvements by the manufacturers, this research examines daily event records, machine recipes, and adjusts the tuning parameters of the machines for optimization. Such adjustments can improve the quality of machine output while increasing production efficiency.

This study proposes a machine learning-based model for predicting the success or failure of tuning parameters for the VARIAN VIISTA HC machine. The dataset used includes a large amount of historical data on semiconductor ion implantation, and the model's accuracy and practicality have been validated through analysis and comparison of experimental data. To enhance the performance of the model, various machine learning algorithms were employed, including decision tree estimation, random forest, XGBoost, and Artificial Neural Network (ANN). Experimental results show that the predictive model achieves an accuracy rate of over 90% on the test set, accurately predicting the success rate of semiconductor ion implantation, and improving production efficiency and product quality through optimized implantation parameters. In this research, a new method was proposed to effectively avoid the range of potential machine failures by limiting the scope of Tune Beam Parameters, thus reducing machine errors. The experimental results confirm the effectiveness of this parameter restriction in improving production efficiency by over 98% accuracy. This research provides an effective production optimization method for the semiconductor manufacturing industry and serves as an example of the application of machine learning in the semiconductor industry. By enhancing competitiveness and exploring new opportunities, this method contributes to achieving long-term business goals.

Keywords: Data mining, Prediction model, Random forest, Machine learning, Ion Implantation.

[1] 林冠名 (2022). "運用LDA建置新冠肺炎假訊息檢測模型" [Using Latent Dirichlet Allocation to Construct a COVID-19 Fake News Detection Model]. 國立成功大學工學院工程科學系碩士在職專班碩士論文.。
[2] 徐琮堡（2015）。運用田口實驗法提升半導體離子植入製程參數設定之效率 [碩士論文，國立高雄應用科技大學，工業與工程管理系碩士 班]。
[3] 袁梅宇. (2016). 王者歸來：WEKA機器學習與大數據聖經[第3版]. 台灣：佳魁。
[4] 郭瑞祥. (2003). 基于K-means算法的半導體製造數據資料探勘研究 [學位論文，國立臺灣大學，台北市，中華民國]. http://ntur.lib.ntu.edu.tw/handle/246246/3078。
[5] A.larson, Lawrence & M.williams, Justin & Current, Michael. (2012). Ion Implantation for Semiconductor Doping and Materials Modification. Reviews of Accelerator Science and Technology. 04. 10.1142/S1793626811000616.
[6] Braha, D., & Shmilovici, A. (2002). Data mining for improving a cleaning process in the semiconductor industry. IEEE Transactions on Semiconductor Manufacturing, 15(1), 91-101. https://doi.org/10.1109/66.983448。
[7] Breiman, L. (2001). "Random Forests." Machine Learning 45(1): 5-32. 。
[8] Breiman, L.,”Random forests,”Machine learning ,45.1, 5-32, 2001.
[9] C. Jasper, A. Hoover, M. Ameen, and F. Sinclair, "A parametric investigation of a high current implanter for long term process performance improvements," in 2000 International Conference on Ion Implantation Technology Proceedings. Ion Implantation Technology - 2000 (Cat. No.00EX432), Alpbach, Austria, 2000, pp. 611-614, doi: 10.1109/IIT.2000.924227.
[10] Chien, C. F., Wang, W. C., & Cheng, J. C. (2006). Data mining for yield enhancement in semiconductor manufacturing and an empirical study. Expert Systems with Applications, 30(4), 713-722. https://www.sciencedirect.com/science/article/pii/S095741740600131X。
[11] Current, Michael. (2012). Ion Implantation for Fabrication of Semiconductor Devices and Materials. 10.1142/9789814307055_0002.。
[12] E.Alpaydin , Introduction to machine learning . Summit Valley press,fourth version ,2020.。
[13] Fawcett, Tom (2006); An introduction to ROC analysis, Pattern Recognition Letters, 27, 861–874.。
[14] G. E. Moore, "Cramming more components onto integrated circuits, Reprinted from Electronics, volume 38, number 8, April 19, 1965, pp.114 ff.," in IEEE Solid-State Circuits Society Newsletter, vol. 11, no. 3, pp. 33-35, Sept. 2006, doi: 10.1109/N-SSC.2006.4785860.。
[15] H. Graoui, A. Al-Bayati and R. Tichy, "Accuracy of doping and process optimization for 0.18 μm PMOS technology," Ion Implantation Technology. 2002. Proceedings of the 14th International Conference on, Taos, NM, USA, 2002, pp. 189-192, doi: 10.1109/IIT.2002.1257970.。
[16] J. Shim, S. Cho, E. Kum and S. Jeong, "Adaptive fault detection framework for recipe transition in semiconductor manufacturing", Computers & Industrial Engineering, vol. 161, 2021.。
[17] J.H. Cha, S.W. Kim and H.J. Lee, "A study on beam extraction characteristics of RF and DC filament ion source for high current ion implanters", Applied Science and Convergence Technology, vol. 30, no. 3, pp. 92-94, 2021.。
[18] moea, “產業經濟統計簡訊”[Online].Available: https://www.moea.gov.tw/Mns/dos/bulletin/Bulletin.aspx?kind=9&html= 1&menu_id=18808&bull_id=10080
[19] Pulipuli (2017, April 25). 不寫程式也能預測未知！用Weka分類模型來預測未知案例 / Make predictions with Saved Machine Learning Model in Weka. [Blog post]. Retrieved from https://blog.pulipuli.info/2017/04/weka-make- predictions-with-saved.html。
[20] Quinlan, J. R. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993.。
[21] rickjung88. (2014). Data Mining 學習路：概念、技術與工具 [Data Mining Learning Path: Concepts, Techniques, and Tools]. IThome. Retrieved from https://ithelp.ithome.com.tw/users/20083470/ironman/733。
[22] S. R. Kurakula and J. Trujillo, "Data mining to detect ion source failures in varian VIISta implanters," 2016 27th Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC), Saratoga Springs, NY, USA, 2016, pp. 149-150, doi: 10.1109/ASMC.2016.7491148.。
[23] Scheuer, J. T., Cucchetti, A., Welsch, M., Callahan, W., Luey, K., & Olson, J. C. (2006). Optimized autotuning for single wafer high-current and medium-current implanters. AIP Conference Proceedings, 866, 381. https://doi.org/10.1063/1.2401536。
[24] Tang, H., & Dunn, M. L. (2011). Carbon implantation performance improvement by (CO) gas co-implantation in silicon. Semantic Scholar. https://www.semanticscholar.org/paper/Carbon-implantation-performance-improvement- by-%28CO%29-Tang-Dunn/a9ea861801263b2611a32a2f5786549f8cb4262a。
[25] U. Fayyad,and P.Smyth,"Form Data Mining to Knowledge Discovery:An Overview," Advances in Knowledge Discovery and Data Mining,pp. 1-36,1996.
[26] Wang, J., Zheng, P., & Zhang, J. (2020). Big data analytics for cycle time related feature selection in the semiconductor wafer fabrication system. Computers & Industrial Engineering, 143, 106362. https://doi.org/10.1016/j.cie.2020.106362。
[27] Xiao, H. (1991). 半導體製程技術導論 [Semiconductor Process Technology Introduction]. (羅正忠, 張鼎張, Trans.). 學 銘圖書有限公司.。
[28] Y. -L. Lin, Q. Zhao and S. -C. Horng, "Machine Learning-Based Approach for Automatic Ion Implanter Monitoring," 2022 International Automatic Control Conference (CACS), Kaohsiung, Taiwan, 2022, pp. 1-6, doi: 10.1109/CACS55319.2022.9969826.。
[29] Quinlan, J. R. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, 1993.