分散式基因演算法應用於資料視覺化

在本論文提出了一個改良式的基因演算法用來處理二維矩陣資料平面視覺化。主要的目的是將一個差異性/距離矩陣的資料平面化成（ｘ，ｙ）座標的形式，以轉換成圖形的方式展現資料之間的差異性，符合人類直觀式的思考，使得複雜的資料能更快速更簡化被解讀。在視覺化的設定部份，本論文設計了一種”三點固定法”，避免平面化後的資料有平移、旋轉的重複情況發生，也將資料的範圍縮小至兩個象限內，大幅降低了基因演算法發生重複計算的可能。在基因演算法方面，本論文加入了動態的基因長度(精準度)變化的想法，使演算法能夠更快的演化到更佳的解，且可套用在不同種類的基因演算法上，因此也比較了不同參數組合的影響，參數包括：演化方法、基因精準度變化方法、突變率、交配率…等等。找出了最佳參數組合之後，將分散式GA中的島嶼模型概念帶入系統中，發展成混合島嶼模型GA，每個島嶼設定不同的生態環境，以增加族群的多樣性，並藉由族群之間個體的遷徙方式，增快演化的速度。

關鍵字

分散式基因演算法；島嶼模型；資料視覺化；視覺化資料探勘；基因演算法

並列摘要

In this paper, we propose an improved Genetic Algorithm to handle data visualization with 2D matrix data. The main purpose is to transform a diversity/distance matrix into a form of (x, y) coordinates, and then we can turn it into a graph in order to show the difference between data. It can be more faster and more reducible to read out the complicated information with the ocular cognition of human thinking. In the setting of data visualization, we design a method which is called "the method with 3 fixed points". It can avoid the occurrence of repeated case with shift and rotation after data visualization. It also reduce the data range to two quadrant and plenty reduce the possibility of reiterative computation of GA. In part of Genetic Algorithm, we promote a function of GA which could evolution faster by dynamically change gene length (accuracy). Furthermore, it can set up on different kinds of Genetic Algorithm and consequently we make a comparison with diverse parameters, such as the way of evolution, gene accuracy, crossover rate and mutation rate... After finding the best combination of parameters, we import the Island Model conception form distributed GA to our system, and promote to Mixed Island Model GA. In order to getting more diversity of populations, we set different ecological environment for each island. Furthermore, we could increase the evolutional speed by migrations of individual between the populations.

並列關鍵字

genetic algorithm ； island model ； data visualization ； visual data mining ； distributed GA

參考文獻

[4] 李捷隆2003分散式基因演算法應用於旅行家銷售員問題，私立中原大學資工研究所碩士論文。

[1] 陳伯銜 2001地形資料視覺化之比較研究，國立台灣大學地理環境資源研究所碩士論文。

[6] J.J.Grefenstette,1986. Optimization of control parameters for genetic algorithms. IEEE-SMC, SMC-16, 122-128.

[8] C.Andrew, M.N.Cara, V.Kaipillil, T.Berwin and G.Ritu, “Statistical Exploratory Analysis of Genetic Algorithms,” IEEE Trans. Actions On Evolutionary Computation, Vol. 8, pp. 405-421, Aug. 2004.

[9] Haupt, Randy L.,1998. Practical genetic algorithms, New York : Wiley Pub.

國際替代計量

分散式基因演算法應用於資料視覺化

未授權

主題瀏覽