鋼鐵產業為導入工業化發展之重要基礎工業,為技術、資本與能源密集度極高之產業,更緊密連結上下游產業。舉凡各民生工業無不與鋼鐵產業息息相關,然而台灣屬於島嶼型國家,雖佔地理優勢,但缺乏天然資源,煉鋼之原料多數依賴進口,隨著國際鐵礦、能源等原料上漲,運輸成本也因油價上漲而提高,導致鋼鐵產業生產成本居高不下。綜觀而言影響鋼品價格因素大致可區分為生產成本、供需關係、市場環境,偋除大環境之影響,生產成本是造成鋼品價格變化的直接因子,但鋼品的供需關係才是真正影響價格走勢的主要因素。因此本研究目的為結合蜜蜂繁殖演算法與自組織映射圖網路,建立一套鋼鐵成品需求漲跌幅預測系統,期望能確切掌握未來的鋼品需求情勢,保護國內鋼鐵市場免於國外的低銷策略傾銷台灣,國內中下游用鋼需求也能藉此得到保障,營造供需雙方互利之局面,使台灣鋼鐵產業更具競爭力。 經實驗結果顯示,本研究所提出的有因素篩選機制之蜜蜂繁殖演化自組織映射圖網路預測系統,優於無因素篩選機制之傳統自組織映射圖網路預測系統,且分割型模式之蜜蜂繁殖演化自組織映射圖網路更能準確預測鋼品需求漲跌幅。可知不同的資料前處理模式,確實會影響鋼品需求預測準確度。
The steel industry is an important basis for industrial development, and it’s characteristics are technology-intensive, capital-intensive and energy-intensive, and closely link with its upstream and downstream industries. Covered the important raw material for livelihood industries. All of those fields are intimately with steel industry. However, Taiwan is the island-oriented country. Although being provided with the geographical advantage, but Taiwan lacks of natural resources. The majority of steel-making raw materials dependent on imports. With rising transportation cost of raw materials trading around the worldwide markets, this situation will trigger the problem of oil price-hike, leading to the production costs of the steel industry stays high. Many factors will affect steel price, including production cost relationship between supply and demand, and market environment; but dismiss the impact of the environment, the production cost is a direct factor of the price fluctuation. More important, the relationship between supply and demand of steel products is the main factors to influence the trend of the involving prices. The purpose of this study is to combine honey-bee mating algorithms and neural network, then establishing a set of demand fluctuation forecast about steel finished goods. We expect the certainty of the future of the steel product demand situation. Not only to protect the domestic steel market from foreign low-priced competitive strategy, but also to guarantee to the domestic steel of supply chain from middle to downstream. Finally, to create ALL-WIN for whole supply chain of the both sides;to be more precisely, making Taiwan's iron and steel industry more competitive. To compared HBMO-based SOM with SOM. The experimental results show that HBMO-based SOM with factors selection is better than SOM with all factors, and the model of divided-HBMOSOM can more accurate the Taiwan steel demand fluctuation. Shows different data pre-processing mode will definitely affect the steel demand fluctuation forecast accuracy.