流行飾品(Costume Jewelry)產業內產品流行性強,商品生命週期短暫,產業進入障礙低。該產業的產業價值鏈多掌握於歐美品牌大廠,業者以中小企業為主並以委託加工為經營策略。為取得企業之競爭優勢,以生產為導向的業者採用設計加工經營模式因運而生,設計加工需以客戶需求為導向,為消弭業者難以取得終端消費者之需求訊息的困境,本研究擬以利用企業內部歷史資料,探討不同區域與通路之間對產品之需求為何。 本研究主要運用資料探勘(Data Mining)中有關集群之技術,透過顧客屬性資料庫、交易資料資料庫與產品屬性資料庫擷取所需要的變數來進行資料探勘的研究。本研究擬採用資料探勘的無監督類神經網路模式,以集群技術的自組織映射圖網路(Self-Organizing Feature Map Network, SOM)的學習模式,建立集群,探索不同區域與通路之間其產品特徵的需求行為。 本研究發現產品類別為產品屬性的重要變數之一,產品類別未進行分類直接採用自組織映射圖網路學習模式,各集群的產品屬性特徵不明顯,無法進行萃取各群體之特徵。產品類別經分類後,於4*4網路拓樸矩陣進行自組織映射圖網路學習模式,各產品類別可區分為四個集群,各集群的主要變數為顧客資料屬性與產品屬性兩類。顧客資料屬性包括顧客區域和顧客類別,產品屬性包括外型、表面飾材顏色和表面鍍層顏色。各集群的產品特徵需求明顯,可幫助研發人員或決策者進行產品設計研發。
Costume jewelry’s fashion cycle is brief and market trends change fast. Also, here is another issue at this field that is low barriers imply that new entrants flood the market. To avoid to trigle price competition, company could creat it’s competitive advantage through oringial design manufacturing. The paper presents to using data mining skill of Self-Organizing Feature Map Network(SOM),to analyze the database including customer profile, transaction and product data. By discovering the valuable information at product innovation, the enterprises would take the advantage of product innovation to solve the issues of brief product cycle and low barriers of entry in costume jewelry industry. According to our research result, we found that SOM’s training should be implanted by functional needs. After determining 4 clusters under different functional product such as Ring, Necklace, Bracelet, Pin and Earring through SOM’s training, each cluster’s characteristics are significant. With the assistance of data mining in oringial design manufacturing, which enhances the manufacturer of costume jewelry competitiveness.