應用基因演算法結合時間序列於台灣地區鋼鐵價格漲跌幅之預測

Applying GA-based Time Series in Predicting the Taiwan Steel Price Fluctuation

李倩瑜(Chien-Yu Lee)

指導教授 ： 邱垂昱

國立臺北科技大學/管理學院/工業工程與管理系碩士班/碩士(2014年)

https://doi.org/10.6841/NTUT.2014.00223

摘要

鋼鐵業有「重工業之母」之稱，亦為國家策略性基礎工業之一，基於鋼鐵工業的發展與其他產業關係緊密相連，並擁有龐大能力帶動中下游產業的發展，影響國家工業化程度及經濟地位，鋼鐵業的發展受到各國政府的高度重視。國內鋼鐵產業相關研究，多配合國家經濟發展與市場供需狀況加以討論，著重於鋼鐵產業現行狀況，因此，本研究期望可以透過建立一套鋼品價格漲跌幅之預測系統，提供鋼鐵業者為採購之評估準則，進而提升台灣鋼鐵產業經營績效，以應付外在的競爭與挑戰。
本研究採用單一屬性移動平均法(MA)與自行建構之單一屬性基因演算法結合加權移動平均法(GA+WMA)、多屬性基因演算法結合移動平均法與權重(GA+MA+W)針對鋼筋與熱軋不鏽鋼捲片做價格漲跌幅之預測，研究結果證實多屬性GA+MA+W預測結果優於其他兩者，兩鋼品預測出的區間準確率分別為53.75%、50.00%，且可透過研究結果觀察各因素對鋼品為正或負相關影響及其影響程度，其中鋼筋受因素長期的影響而熱軋不鏽鋼捲片受因素短期的影響。

關鍵字

預測 ； 基因演算法 ； 移動平均法 ； 加權移動平均法

並列摘要

The steel industry is known as the mother of heavy industry and is one of the strategic basic industries. Based on the relationship between the development of the steel industry and other industries are closely connected and the steel industry has a huge capacity to promote the development of the middle and lower stream of industries. It affects the degree of industrialization and a country’s economic status. Every country quite values the development of steel industry and regard as the important development projects. The researches of domestic steel industry are involved with a country's economic development and supply and demand on the market. Therefore, it mostly focus on the current situation of steel industry. Therefore, via constructing the prediction systems, the findings of the study can show the prediction of the price fluctuation of steel products and provide the steel industry evaluated criteria of purchasing and improve the overall operating performance of Taiwan's steel industry to cope with external competition and challenges.
This study uses a single attribute moving average (MA) and via constructing a single attribute Genetic algorithm combines with the weighted moving average (GA+WMA), multi-attribute Genetic algorithm combines with the moving average and the weight (GA+MA+W) to predict price fluctuation of rebar and hot rolled stainless coil. The study results confirm multi-attribute GA+MA+W prediction outperforms the others. The accuracy rates of the price fluctuation interval for the two steel products are 53.75%, 50.00%, respectively. Via the results, we can observe various factors on steel products which is positive or negative correlation and the level of its impacts. The factories affected the rebar price fluctuation consistently, the factories affected the HR stainless coil price fluctuation transitorily.

並列關鍵字

predict ； Genetic Algorithms ； Moving average ； Weighted Moving average

參考文獻

1. 林詩彥，鋼鐵價格決定機制及影響因素分析，碩士論文，中原大學國際貿易研究所，桃園，2006。

21. 溫珮玲，散裝海運市場運價決定機制及影響因素分析，碩士論文，中原大學國際貿易研究所，桃園，2005。

13. 翁怡欣，應用蜜蜂繁殖演化結合自組織映射圖網路於台灣地區鋼鐵價格漲跌幅之預測，碩士論文，台北科技大學工業工程與管理系，台北，2013

14. 施柏州，應用蜜蜂繁殖演化倒傳遞類神經網路於台灣地區鋼鐵價格之預測，碩士論文，台北科技大學工業工程與管理系，台北，2012

6. 謝明瑞，台灣鋼鐵業的發展，國政研究報告，財團法人國家政策研究基金會，台北，2001。