利用遺傳演算法作最佳化分群

Abstract

資料分群法是一種複雜性的最佳化問題，它的應用所涵蓋的範圍從聲音和影像處理到資料傳送和儲存等等。我們討論一個分群法結合Fuzzy c-means演算法則，Fuzzy c-means初值(initial seed point)，失真誤差(distortion error)，複雜度成本(complexity cost)，以及遺傳演算法(Genetic Algorithm)。我們的演算法利用傳演算法和成本分析（含失真誤差及複雜度成本）決定Fuzzy c-means分群法的初值，及決定Fuzzy c-means群數；模擬結果顯示這個新的演算法能得到較佳的結果。

Data clustering is a complex optimized problem with applications ranging from speech and image processing to data transmission and storage in technical as well as in biological systems. We discuss a genetic clustering algorithm that jointly optimizes fuzzy c-means algorithm, initial seed points, distortion errors, complexity cost and genetic algorithm. Agenetic algorithm and cost function (i, e, complexity cost and distortion cost) are used to determine the initial seed points and number of clusters for fuzzy c-means clustering algorithm. Experimental demonstrate that the algorithms can reach to the optimized or near-optimized solution.