產品品質是企業追求客戶滿意、提高競爭力的重要指標，生產檢測則是確保產品品質的重要 手段，然為節省成本及時間，目前業界採取抽檢方式來進行品質監測，但如此並無法確保每一個 產品品質，故透過資料分析手法來改善工廠之生產品質與效率是一重要課題。本研究所提出技術 採用機器學習方式進行大數據資料建模，包含支援向量機、邏輯斯迴歸及決策樹等各式分類器， 在普遍製造現場資料特性未知情形下，以袋式(Bagging)整體學習(Ensemble Learning)機制進行模 型訓練，有效降低傳統方法因選擇錯誤分類器而造成過適(overfitting)風險，達成穩健的品質分類 效能，以低時間延遲方式，檢測產品的品質與良率，當不良品數目過高或良率不佳時，警示現場 作業人員進行必要的停機檢查。導入本技術可幫助現場人員即時掌握在製品品質狀況，防杜廢品 進入下一到製程而產生之二次損失成本，同時確保產品品質。

In order to gain customer satisfaction and enhance competitiveness, manufacturers usually aim to offer their clients with high quality products. However, under the consideration of time and cost reduction, random sampling has become a popular method for production quality metrology whereas not all the products will be examined and assured. To overcome this problem, this paper proposes a WIP Quality Detection and Classification method based on Ensemble Learning with well-known machine learning approaches, like Support Vector Machine (SVM), Logistic Regression, Decision Tree, and so on. Unlike a statistical ensemble in statistical mechanics, which is usually infinite, a machine learning ensemble refers only to a concrete finite set of alternative models, but typically allows for much more flexible structure to exist among those alternatives. Therefore, while most of the manufacturing data characteristics are unknown (means infinite hypothesis), our method can reduce traditional model bias from overfitting to enable real-time product quality detection with achieving stable classification efficiency. Once the yield rate is below cert ain level, the system will send a warning notice to inform the process engineer to shut down the process and cut the costs of rework, such as correcting of defective, failed, or non-conforming item, during or after inspection.