Full metadata record
DC FieldValueLanguage
dc.contributor.author陳俊詠zh_TW
dc.contributor.author黃冠華zh_TW
dc.contributor.authorChen, Jyun-Yongen_US
dc.contributor.authorHuang, Guan-Huaen_US
dc.date.accessioned2018-01-24T07:35:06Z-
dc.date.available2018-01-24T07:35:06Z-
dc.date.issued2016en_US
dc.identifier.urihttp://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070352621en_US
dc.identifier.urihttp://hdl.handle.net/11536/138370-
dc.description.abstract機台生產資料的分析近年來在產業界相當熱門,好的分析及預測將會使得問題被發現更甚至可降低生產成本或提升良率。本文將使用MOCVD機台的生產資料利用線性混合效用模型(linear mixed effect model)進行資料的分析與預測(prediction),目的為預測機台中過濾器(Particle Filter)的損壞。我們將會從如何整理原始資料開始,並對資料做特徵值萃取,再經由變數選擇(Variable Selection)的方式選取出有用的特徵值。變數選擇使用的工具有F檢定、概似比檢定、AIC、BIC、Conditional AIC及Lasso。最後使用交叉分析(cross-validation)的方式來比較各種變數選擇後的模型的好壞,並針對其結果做出合理的解釋,更可以藉由我們挑選出的最終模型對於此類型機台零件做出合理的損壞預測,以提供業界一個可行的預測方式並可以給予此類型資料的研究者做為未來研究的參考。zh_TW
dc.description.abstractAnalyzing the data generated in the production process can help us identify problems, and then reduced the budge and improve the yield rate. The main purpose of this thesis is to predict the damage of particle filter component in the MOCVD machine. We mainly use linear mixed effects models to analyze the data generated from the sensors in the MOCVD. The procedure starts with data mining, and uses various variable selection method, including F-test, likelihood ratio test, AIC, BIC, Conditional AIC and Lasso, to determine the key features. Then, the cross-validation is used to select the best variable selection method. Finally, we show that the prediction from the final model is a reliable reference for life time forecasting.en_US
dc.language.isozh_TWen_US
dc.subjectMOCVD機台zh_TW
dc.subject過濾器zh_TW
dc.subject變數選擇zh_TW
dc.subject線性混合效用模型zh_TW
dc.subject預測zh_TW
dc.subjectF檢定zh_TW
dc.subject概似比檢定zh_TW
dc.subjectAICzh_TW
dc.subjectBICzh_TW
dc.subjectConditional AICzh_TW
dc.subjectLassozh_TW
dc.subject交叉分析zh_TW
dc.subjectMOCVDen_US
dc.subjectParticle Filteren_US
dc.subjectvariable selectionen_US
dc.subjectlinear mixed effect modelen_US
dc.subjectpredictionen_US
dc.subjectF-testen_US
dc.subjectlikelihood ratio testen_US
dc.subjectAICen_US
dc.subjectBICen_US
dc.subjectConditional AICen_US
dc.subjectLassoen_US
dc.subjectcross-validationen_US
dc.title使用線性混合效用模型於MOCVD機台過濾器零件之資料分析及損壞預測zh_TW
dc.titleLinear Mixed Effects Models for Data Analysis and Error Prediction of MOCVD Particle Filter Componentsen_US
dc.typeThesisen_US
dc.contributor.department統計學研究所zh_TW
Appears in Collections:Thesis