在汽車銷售市場逐年萎縮，各車廠在製造端沒有獲利的情況下，紛紛將目標轉為仍有廣大需求的售後服務市場，並盡力降低零件呆料，以提升企業競爭力。 依據歷史經驗，庫存之零件變成呆料之可能性會隨著庫存年限的增加而變高。但未異動ㄧ年之內的零件，通常能藉著促銷或是折扣之方式進行銷售，而避免變成呆料。所以如何能有效的挑選出容易成為呆料之零件(即滯銷零件)，使管理者能提前進行促銷或是進行備料量之調整，便成了此次研究想探討的問題。 本研究以C公司為例，收集大量庫存零件資料，並利用資料探勘手法，配合Clementine軟體，建立三種模型，其為羅吉斯迴歸預測模型、羅吉斯迴歸預測模型-c合併變數模型及C&RT決策樹模型，經過三項模型比較後，得知以C&RT決策樹建立預測模型，可有效預測呆料風險，進而降低呆料發生機率，可作為後續各車廠降低成本之參考。 In the downturn of the automobile market and under the situation of earning a little profit, those automobile manufacturers shift their target to the after-sales markets, which still have big demand. Besides, In order to improve their company competition, they also try their best to reduce dead parts quantity. According to the experience, the dead part probability of inventory stock will increase with the inventory years. If the part has sale history in one year, it often can avoid becoming dead part by using promotion or discount. So, how to pick up the high risk dead part and let manager to do promotion or discount in advice is this thesis main purpose. This thesis is using C company for research data. Besides, the thesis also uses data mining methodology and Clementine software to build three prediction models. They are Logistic Regression, Logistic Regression with PCA/Factor and C&RT decision tree. After evaluating these three models, this thesis recommends using C&RT decision to prediction model. It can decrease the risk of dead parts efficiently, and it can be a reference for motor company.
