在台灣紡織產業中，瑕疵的發生是一個不可避免且持續存在的問題，在生產過程中，機器參數的設定影響到最終產品的良率，如果設置不當導致產品出現各種類型的瑕疵，不僅影響產品的品質，還會導致大量原料浪費，增加了生產成本和資源消耗。 如何有效識別出具有瑕疵相依性之參數是本研究之重點，本研究根據合成資料集與紡織資料集中使用多變量迴歸樹，找出對瑕疵有影響之特徵集，辨識出每個關鍵特徵的交集範圍，並且將各關鍵特徵之範圍組合成規則。 本研究將所提出的方法與先前的研究方法對合成資料集與紡織資料集進行分析與比較。合成資料集的實驗結果顯示本研究方法有效提升了18% 準確度、整體的執行時間快先前的研究方法2倍。紡織資料集的實驗結果，本研究所產生的規則為狹窄且精確的參數範圍，更適合提供建議給操作人員。以上實驗結果表示本研究方法能夠有效識別具有瑕疵相依性的資料集中參數的交集範圍。;The textile industry in Taiwan, the occurrence of defects is an inevitable and ongoing problem. During the production process, the settings of machine parameters affect the yield of the final product. Improper settings will lead to various types of defects in the product. It not only affects the quality of the product, but also leads to a large amount of waste of raw materials, increasing production costs and resource consumption. How to effectively identify parameters with defect dependencies is the focus of this research. This research uses multiple response regression tree based on synthetic datasets and fabric datasets to find feature sets that have an impact on defects, identify the intersection of each key feature, and combine the range of each key feature into rules. This study analyzes and compares the proposed method with previous research method on synthetic datasets and fabric datasets. Experimental results on synthetic datasets show that the proposed method effectively improves the precision by 18%, and the overall execution time is 2 times faster than previous research method. Experimental results on fabric datasets show that the rules generated by the proposed method have a narrow and precise parameter range, which is more suitable for suggestions to operators. The above experimental results indicate that the proposed method is effective in identifying the intersection range of parameters in datasets with defect dependencies.
