整合基因演算法及類神經網路於印刷電路板生產預測之研究

A Study of Integrating Genetic Algorithm & Neural Network In Forecasting Problem of Printed Circuit Board Industry

楊東翰(Tung-Han Yang)

指導教授 ： 張百棧 博士

元智大學/工程學院/工業工程與管理學系/碩士(2004年)

摘要

印刷電路板產業在我國經濟中，扮演相當重要的角色。但目前的印刷電路板產業普遍存在庫存堆積與部分停工待料的問題。然而，透過一套準確的需求方式可以有效減少這類問題發生的機率與所造成的成本損失。
因此，本研究提出一套結合基因演算法(Genetic Algorithm)與類神經網路(Neural Network)之預測方式。利用灰色關聯分析從眾多的相關因素中遴選出關聯度最高的因素，並利用溫氏指數平滑法考量趨勢性因素與季節性因素所造成的影響。最後將上述各種因子之數值與過去之實際需求資料放入基因演算法結合類神經網路中進行訓練，並以平均絕對百分比誤差、平均絕對偏差量與偏差成本與其他預測方法進行比較。
實驗結果顯示，基因演算法結合類神經網路較傳統的統計模型與倒傳遞類神經網路預測準確，可以提供相關產業進行需求預測時之參考。

關鍵字

生產預測 ； 印刷電路板 ； 基因演算法 ； 類神經網路 ； 灰關聯分析 ； 溫氏指數平滑法

並列摘要

Printed circuit board industry plays an important role in our nation’s economy, but severe inventory stacking and material lacking problems still exist. However, it is likely to decrease the probability of occurrence and reduce the cost of these kind problems via establishing an accurate demand forecasting system.
Thus, a forecasting model integrating Genetic Algorithm and Neural Network (GANN) has been proposed in this research. Along with trend and seasonal factors considered by Winter’s Exponential Smoothing method, effective economical factors are chosen by the Grey Relation Analysis. The numerical data of these factors and actual demand of past 5 years are inserted into training stage of GANN, while the comparison with other models is evaluated on testing stage in terms of Mean Absolute
Percentage Error (MAPE), Mean Absolute Deviation (MAD) and Deviation Cost.
The result of experiments shows that the performance of GANN is superior to traditional statistical models and Back Propagation Network. The GANN provides a splendid solution to the forecasting problem for relevant industries.

並列關鍵字

Forecasting ； Printed Circuit Board ； Genetic Algorithm ； Neural Network ； Grey Relation Analysis ； Winter''s Exponential Smoothing

參考文獻

1.Abdel-Aal, R.E, Al-Garni, A.Z, “ Forecasting monthly electric energy consumption in eastern Saudi Arabia using univariate time series analysis, ” Energy, Vol. 22, pp. 1059-1069, 1997.

2.Arifovic, J. R. Gencay, “ Using genetic algorithms to select architecture of a feedback artificial neural network,” Physica, Vol. A, pp.574-594, 2001.

3.Chow, T. W. S. and C. T. Leung, “Nonlinear autoregressive integrated neural network model for short-term load forecasting,” IEE Proceeding Online no. 19960600, pp500-506, 1996.

4.Fausett, L., Fundamentals of Neural Network, Prentice Hall International, 1994.

5.Hsu, P. H., C. H. Wang, Joseph Z. Shyu and H.C. Yu, “A Litterman BVAR approach for production forecasting of technology industries,” Technological Forecasting & Social Change, Vol.70, pp.67-82, 2002.