A New K-Means Clustering Based on Genetic Algorithm

Abstract:

The driving force of a genetic algorithm is the fitness function. The traditional fitness function based on the error sum of squares has a poor effect when the number of difference between classes is huge. Furthermore, it often fails to get better clustering results due to the lack of the clustering effect evaluation. So, this paper proposes a new K-means clustering based on Genetic Algorithm. Firstly, we define the concept of difference density and uses it as the fitness function of the genetic algorithm. Secondly, we evaluate the quality of the clustering results by some indicators and find the best cluster centers by iteration. The comparison proves that our method achieves a better result than the traditional K-means clustering based on genetic algorithm.

• Periodical: Applied Mechanics and Materials (Volume 203)
• Pages: 94-98
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.203.94
• Online since: October 2012
• Authors: Chang Sheng Cheng, Yan Meng Shang
• Keywords: Difference Density, Genetic Algorithm (GA), K-Mean, Quality of Clustering

References

[1] Goldberg D.E. Genetic Algorithms in Search, Optimization and Machine Learning. New York: Addison-wesley(1989).

[2] Xiaoping Wang, Liming Cao. Genetic algorithm - the theory, application and software implementation. Xi'an: Xi'an Jiao Tong University press(2002).

[3] Xiangyang She, Huifeng Xue, Wudi Xu. Multidimensional fast clustering based on genetic algorithm. Application Research of computers. 06(2005).

[4] Jingguang Fu, Gang Xu, Yuguo Wang. Clustering analysis based on genetic algorithm. Computer Engineering. 04(2004).

[5] Yan Yang, Fan Jin, KAMEL Mohamed. Evaluation validity of clustering. Application Research of computers. 06(2008).

[6] Yanchi Liu. The method for determining t the number of clustering based on density, China management informationization. 09(2011).