應用智慧型灰關聯於國小模範生選拔

Applying intelligence of Grey Relational for the selection of model students

劉華奇(Hua-Chi Liu)

指導教授 ： 施能義

亞洲大學/資訊學院/資訊工程學系碩士班/碩士(2008年)

摘要

國小模範生選舉一直以來都是國小重要的活動，而選舉標準及選舉結果很難讓大家都滿意，如果有一套方法能減少人為因素的影響，相信結果能讓大家都接受。本研究提出智慧型灰關聯分析來選拔模範生，以減少這些問題。
而進行灰關聯分析中必須面對的一個問題，就是事先建立一個目標準則來選擇ρ值，爾後才能對事物之間的關聯度依此準則下進行討論。本研究以全校模範生為標準參考，其他模範生候選人作為比較，並提出以具有最大關聯度間的平均差距，即關聯度間隙的最大分散性來選取分辨係數ρ，且以遺傳演算法作為選取運算方法，使得本研究所提方法具有智慧型通用性之價值。研究結果，得到兩個結論：
1.灰關聯度描述系統發展中，變數間相對變化的狀況，也就是說變化大小、方向與速率等的相對性，如果兩個數列間，相對應變化是一致時，則認為兩者關聯度大。而在灰關聯分析中，涉及到其中一個重要的分辨係數的取值問題，尤其不同分辨係數必然導致灰關聯度的排序的不同。因此，在灰關聯度分析中應有一個比較準則作為選擇分辨係數的依據。本研究則提出具有最大灰關聯度分散性的具體方法，並經實際範例模擬說明，除排序上具有與其他方法的一致性外，並有最大分辨的平均灰關聯度間距，使分辨係數的要求具有客觀性。雖然在要求準則的目標函數為二元二次方程，但本研究引進遺傳運算，只要求所解決的問題是可計算的，不需可微分性及其它要求就可得到分辨係數，使得本研究所提方法具有很大的使用範圍與通用性。
2.利用本研究所提出的模型，將可以預測出與目前班級選舉選拔出同一個模範生，而且本模型具有智慧型，鍵入所有數據後，就可預測模範生，未來可以利用此一模型來預測模範同學的塑造，適合任一班級來使用，減少許多麻煩問題。

關鍵字

模範生 ； 灰關聯分析 ； 分辨係數 ； 分散性

並列摘要

Though the criteria and the results of the model students’ selection can’t be satisfied by everyone, it has been an important activity in primary schools. If there is a more equitable solution to avoid human control, thus the results will be more acceptable for everyone. The purpose of this research is to solve this problem by applying the intelligence of grey relation analysis to select model students.
In order to carry the grey rational analysis, researcher has to set up a goal criterion to choose the ρ value beforehand. According to this criterion, further discussion will be able to be carried. Researcher use all the model students as the frame of reference, the other candidates as the relative reference, and present the average disparity that has the biggest relation interval, namely, the biggest dispersion of the relation interval, to select the discriminating coefficient ρ. Genetic algorithm is used to operate the selection. The results indicate that:
1. In grey relation grade description system development, variable relative change condition, in other words changes the size, the direction and the speed and so on the relativity, when during two sequences, relatively should change is consistent, then thinks two connections big. But in the grey relation analysis, involves to important distinguishing coefficient value question, the especially different distinguishing coefficient inevitably causes grey relation grade arrangement the difference. Therefore should have a comparison criterion in grey relation grade analysis to take the choice distinguishing coefficient the basis. This study proposes the biggest dispersion of the relation, to select the discriminating coefficient ρ. Examples in this study prove that the value of the coefficient ρ is objective. Besides, genetic algorithm is used to operate the selection in this study, the distinguishing coefficient α can be got without the need of differentiation. Thus, the method proposed in this study can be used widely.
2. This study proposes a method that can predict the right model student. And the result is just the same as the election in his class. After inputting all the data of candidates, this method can predict the right model student without any human control or other factors. In the future, this method can be used to predict the model student in any class.

並列關鍵字

Model students ； Grey relation analysis ； Discriminating coefficient ； Dispersion

參考文獻

[26] 林豐澤，演化式計算下篇：基因演算法以及三種應用實例，智慧科技與應用統計學報，第三卷，第一期，pp.29-56，2005年6月。

[25] 林豐澤，演化式計算上篇：演化式演算法的三種理論模式，智慧科技與應用統計學報，第三卷，第一期，pp.1-28，2005年6月。

[24] 陳昭吉，應用遺傳基因演算法於晶圓製造場之設施佈置問題，元智大學工業，工程與管理研究所碩士論文，2003年。

[33] Holland, J. H., Adaptation in Natural and Artificial Systems,Ann Arbor, MI: The University of Machigan Press, 1975.

[34] Coello, C. A. C.,“A Comprehensive Survey of Evolutionary-based Multiobjective Optimization,” Knowledge and Information Systems, 1999.