應用基因演算法結合時間序列於台灣地區鋼鐵需求漲跌幅之預測

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

鄭佩君(Pei-Chun Cheng)

指導教授 ： 邱垂昱 ； 簡禎富

國立臺北科技大學/管理學院/工業工程與管理研究所/碩士(2014年)

https://doi.org/10.6841/NTUT.2014.00033

摘要

鋼鐵產業的發展受全球市場的影響，屬於經濟與工業發展的基礎，雖然造成鋼鐵價格變化的直接因子是生產成本，但供需關係才是影響價格走勢之重要因素。因此國內鋼鐵產業首要任務，即是維持鋼鐵之供需平衡，唯有確切掌握鋼鐵未來之情勢分析，才能保護國內鋼鐵產業免於國外的低銷策略，而國內中下游之用鋼需求也才能藉此得到保障，並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，本研究所提出的多重屬性基因演算法結合移動平均法與權重之預測模型較優於單一屬性移動平均法及單一屬性基因演算法結合加權移動平均法之預測模型，並提升4%至5%之預測誤差；而在十二期多重屬性基因演算法結合移動平均法與權重之實驗結果與文獻之漲跌幅區間準確率比較後，在熱軋不銹鋼板捲與鋼筋時，本研究之區間準確率皆為45.00%，且優於文獻之實驗結果。可知以線性方式計算鋼品需求漲跌幅之準確率，進而提升鋼品需求之區間準確率。

關鍵字

預測 ； 最佳化 ； 移動平均法 ； 加權移動平均法 ； 基因演算法

並列摘要

Development of steel industry affected by global markets, are the basis of economic and industrial development, although a direct factor causing changes in steel prices is the cost of production, but the relationship between supply and demand is an important factor affecting the price trend. Therefore, the primary task of the domestic steel industry, which is to maintain the supply-demand balance of steel, the only certainty on the future of the steel situation analysis in order to protect the domestic steel industry from low marketing strategy abroad, while middle and lower reaches of the domestic demand for steel also can borrow this is guaranteed, and create mutual benefits of both supply and demand situation, so that Taiwan's steel industry more competitive. The experimental results show that our proposed multi-attribute genetic algorithm combined with the weight of the moving average method is superior to the right than the prediction model prediction model and a single attribute moving average single attribute genetic algorithm combined with the weighted moving average method, and enhance 4 % to 5% of the prediction error; while in the second phase of multiple properties combined with the genetic algorithm results and weight of the moving average method right after the interval and Price comparison accuracy literature, in the hot-rolled stainless steel coils and steel, the interval study are 45.00% accuracy rate, and superior results literature. Shows that the accuracy rate is calculated in a linear fashion up the decline of demand for steel products, thereby enhancing the accuracy range of steel products demand.

並列關鍵字

predict ； moving average ； weighted moving average ； Genetic Algorithm

參考文獻

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

38.林旭信，都市雨水下水道系統最佳化操作模擬，博士論文，臺灣大學土木工程學研究所，台北，2005。

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

32.林豐澤，演化式計算下篇：基因演算法以及三種應用實例，文化大學應數系智慧科技與應用統計學報， 2002，pp. 29-56。

29.魏大惟，契約價值差價交易、共整合、與移動平均，碩士論文，朝陽科技大學財務金融系碩士班，臺中，2011。