模糊c-均數演算法 (Fuzzy C-Means Algorithm , FCM Algorithm) 是聚類分析常用的方法。當資料中存在著干擾變數 (noise variables) 時，模糊c-均數演算法之錯誤率相對提高。如何選取變數之權數以降低錯誤率是一重要課題。基於此，本文提出了一種新的變數權數選取方法，稱為共變異矩陣（Covariance Matrix, CM）方法。由模擬結果顯示：所提之變數選取方法能有效降低分群之錯誤率。最後，將所提之 CM 方法應用於彩色影像分割。


關鍵字: 模糊c-均數演算法、變數權數、共變異矩陣、彩色影像分割。

Fuzzy c-Means Algorithm (FCM Algorithm) is a commonly used method of clustering analysis. When there are noise variables in the data, the error rate of the fuzzy c-means algorithm is relatively improved. How to choose the weight of the variable to reduce the error rate is an important issue. Based on this , this paper presents a new method of variable weight selection, called Covariance Matrix (CM) method. The simulation results show that the proposed variable selection method can effectively reduce the error rate of clustering. Finally , the proposed CM method is applied to color image segmentation.


Keyword: Fuzzy C-Means Algorithm、feature-weight、Covariance Matrix、color image segmentation.

Contents

Abstract Ⅰ
摘要 Ⅱ
Contents Ⅲ
List of figures Ⅳ
List of tables Ⅴ

1. Introduction 1

2. proposed approach to feature-weight selection 2

2.1. Motivated example 2

2.2.1. Weighted Fuzzy C-Means algorithm 2

2.2.2. Weighted Fuzzy C-Means algorithm steps 3

3. Experimental comparisons 4

3.1. Proposed feature-weight 4

3.2. Feature-weight of motivated example 5

3.3. Feature-weight of Iris Data 8

3.4. Combined with WFCM and feature-weight in the color image segmentation 10

4. Conclusion 18

References 19

References

Basak, J., De, R.K., Pal, S.K., 1998. Unsupervised feature selection using
a neuro-fuzzy approach. Pattern Recognition Lett. 19, 997–1006.

Fisher, R., 1936. The use of multiple measurements in taxonomic
problems. Ann. Eugenics 7, 179–188.

Hung Wen-Liang., Miin-Shen Yang., De-Hua Chen.,2008. Bootstrapping approach to feature-weight selection in fuzzy c-means algorithms with an application in color image segmentation. Pattern Recognition Letters 29 (2008) 1317–1325.

Hung Wen-Liang., De-Hua Chen.,Jenn-Hwai Yang.,2016.Weighting
variables in Kohonen competitive learning algorithms, Journal of Applied Statistics, DOI:10.1080/02664763.2016.1168367.

Kim, D.W., Lee, K.H., Lee, D., 2004. A novel initialization for the fuzzy
c-means algorithm for color clustering. Pattern Recognition Lett. 25,
227–237.

Modha, D.S., Spangler, W.S., 2003. Feature weighting in k-means
clustering. Machine Learn. 52, 217–237.

Pal, S.K., De, R.K., Basak, J., 2000. Unsupervised feature evaluation: A
neuro-fuzzy approach. IEEE Trans. Neural Networks 11, 366–376.

Wang, X.Z., Wang, Y.D., Wang, L.J., 2004. Improving fuzzy c-means
clustering based on feature-weight learning. Pattern Recognition Lett.
25, 1123–1132.

Yu, J., Cheng, Q.S., Huang, H.K., 2004. Analysis of the weighting
exponent in the FCM. IEEE Trans. Systems Man Cybernet. 34, 634–
639.

Yu, J., Yang, M.S., 2005. Optimality test for generalized FCM and its
application to parameter selection. IEEE Trans. Fuzzy Syst. 13, 164–
176.

Zadeh, L.A., 1965. Fuzzy sets. Inform. Contr. 8, 338–353.