3C電子產品已經是現今人類密不可分的一項產品之一，然而在各廠家的產品價格的競爭，廠商的採購價格錙銖必較，而供應商如何有效的取得既有的毛利外，如何透過人工智慧的方式，預測出建議售價或數量級距的售價趨勢，是目前產業上的研究方向之一，本研究以供應商的角度，預測相關零組件的價格趨勢。 台灣是工業電腦品牌廠商匯聚的一個區域，隸屬工業電腦範疇的上市櫃公司起碼30家以上，工業電腦應用範圍甚為廣泛，但生產規模不及一般商用電腦的數量龐大，在產品價格上一直維持以少量多樣變化，高毛利的準則，讓工業電腦廠商保持有充沛的盈餘。然而在工業電腦設備的最高單價零件，屬中央處理器(CPU)，故在採購上就會有錙銖必較的現象，提升毛利率的重要指標之一。 隨著COVID-19疫情爆發，首當其衝的運輸產業頓時進入了寒冬時期與工廠停工影響，導致在科技產業上的衝擊甚大。影響最前線的半導體產業，受停工規定與人員移動限制，產品出貨的短缺，整體供應鏈斷絕，導致市場上電子零組件現貨價格紛紛調漲。然而在工業電腦產業中，衝擊著少量多樣的生意型態。本研究主要針對CPU的市場商用(Desktop)現貨價格趨勢與工業用(Embedded) CPU現貨價格趨勢以運用長短記憶型類神經網路 (LSTM)和門控遞迴單元 GRU (Gated Recurrent Unit) 預測價格方法參考比較。進而分析LSTM模型和GRU模型在預測之準確度及LSTM和GRU模型之效益。其預測準確率作為供應商銷售的參考指標，爭取供應商該有的毛利。 ;Abstract 3C electronic products are already one of the inseparable products of today′s human beings. However, in the competition of product prices of various manufacturers, the purchase price of various manufacturers must be compared, and how suppliers can effectively obtain the existing gross profit, How to predict the price trend of the suggested selling price or the order of magnitude through artificial intelligence is one of the current research directions in the industry. This research predicts the price trend of related components from the perspective of suppliers. Taiwan is an area where industrial computer brand manufacturers gather. There are at least 30 listed cabinet companies in the industrial computer category. Industrial computers have a wide range of applications, but the production scale is not as large as that of general commercial computers. A small amount of diverse changes and the principle of high gross profit allow industrial computer manufacturers to maintain a sufficient surplus. However, the highest unit price in the industrial computer is the central processing unit (CPU), so there will be a phenomenon of paying for money in the purchase, which is one of the important indicators to improve the gross profit rate. This research mainly focuses on the spot price trend of commercial (Desktop) CPU and industrial (Embedded) CPU spot price trend in order to use long short-term memory type neural network (LSTM) and gated recurrent unit GRU (Gated Recurrent Unit) to predict the price. Method reference comparison. Then analyze the accuracy of LSTM model and GRU model in prediction and the benefit of LSTM and GRU model. Its forecast accuracy rate is used as a reference indicator for supplier sales, and strives for the gross profit that suppliers should have.
