在液晶面板Array製程中，電性檢測的資訊僅提供參考，並未列入報廢或是修補的依據。目前後段Cell製程的品質分類則需要經過點燈檢查才能夠確定，現階段並沒有前製程Array段與後製程Cell段之間的品質模式。因此，若能夠探勘出Array段中的電性量測值與Cell段判定品質評估關係，將可有效降低生產製造成本。本研究應用羅吉斯回歸進行分析，找出電性檢測的關鍵檢測量測值與經面板點燈檢查（P檢）後判定品質異常的預測模型。以臺灣北部某面板製造商所提供的資料進行實證探討，分析結果可得到VTH、RSL、RGE2、RSD1、RSD2為關鍵電性量測值，以及Cell段的品質預測模型，經測試該模型有96.17%的正確分類率。研究成果經評估其具體效益為使公司每個月平均節省840,600元與45小時的製程時間。本研究成果可提供面板製造廠商在制訂前製程所要進行品質策略時的重要參考依據。

In the LCD manufacturing process, the TEG (Test Element Group) electric measure value on Array's process and Array NG Rank are highly correlated in the Cell process. The panel is judged as Array NG Rank 4 will be scrapped. If the Rank in Array NG can be predicted, the waste and the cost can be reduced from unnecessary materials be input into the scrapped panels. Therefore, the research developed a defects rank prediction model by analyzing electric character on Array's process. The model development is divided into two parts. Firstly, using the Logistic regression analysis to find out a better predictive rate in predicting Array NG's Rank and introduce key factors from a group of TEG values. Secondly, according to the analysis results, the models were evaluated based on the predictive accuracy rate of the test sets. The Logistic regression model had 96.17% classification correction rate and the key factors are Vth, Rs1, Rge2, Rsd1 and Rsd2. The prediction model is constructed by analyzing the correlation between TEG electric measure values and Array NG Ranks. The results showed that there is a high ability (96.15%) to predict Array NG Rank 4. The cost saving reaches NT 840,600 dollars with 45 work hours per month.