張淳智黃彥翔2025-08-282025-08-282013-07-09U0061-1408201313301100https://www.airitilibrary.com/Publication/alDetailedMesh1?DocID=U0061-1408201313301100https://nutcir-lib.nutc.edu.tw/handle/123456789/1185本研究主要目的是發展台灣製造業供應鏈完善的分群與分層之架構，並建構完整的層次類型模型。首先透過因素分析篩選變數，再經由兩階段分群法與自組織映射網路進行群集分析，並利用C5.0演算法進行分類，藉以找出更正確的分群，將廠商分為四群：「外銷流動性較差型」、「內銷流動性較好型」、「外銷流動性較好型」、「內銷流動性較差型」。接著利用倒傳遞網路萃取各群之關鍵影響因素，最後使用存貨掌控度進行層次分析，並對關鍵影響因素與存貨掌控度給予評分，再歸納以上實驗結果提出完整的群與群轉換及層與層提升之改善方針。This study aimed to develop a framework of clustering and category model for manufacturers in Taiwan. First, we screen variables through factor analysis, and applied two phases clustering method and self-organizing map for cluster analysis. Then we used C5.0 to evaluate the better clustering method and produced the classification rules. Through these rules, the firms were divided into four groups: foreign trade and weaker liquidity type, domestic sales and good liquidity type, foreign trade and good liquidity type, and domestic sales and weaker liquidity type. Following we used Back Propagation Network to extract the key influence factors of Informed inventory time and Inventory Inaccuracy for each group, and then calulated the score of Informed inventory time, Inventory Inaccuracy, and their key influence factors for each group and category. Finally, this research summarized the above results and provided full path conversion strategies for firms which want to transfer to another group or category.zh供應鏈管理自組織映射網路C5倒傳遞網路分類與分群Supply Chain ManagementSelf-Organizing MapC5Back Propagation NetworkClustering and Classifyingmaster thesis