414 Kyoung-jae Kim and Hyunchul Ahn Table 2. The result of Chi-square analysi SOM GA K-Means Variables Chi-square Asy. Chi-square Asy. Chi-square Asy. Value Sig Value 17.8670.0222 16.44700364 14.5980.0670 000 0.0000 Income 0.0000 134.269 0.0000 Education 0.0002 72.062 0.0000 Occupation 1273470.0000 137.1150.0000 136.3040.0000 Years of web 1974160.00002030040.00002063840.0000 出 528.6720.00005358880.00005594810.0000 urchases 10279100.00001003.7250.000010179090.0000 over the Web eb or 62.9290.00006540130.0000654.8690.0000 5 Concluding Remarks In this study, we propose new clustering algorithm, GA K-means, and applies it to the market segmentation of electronic commerce. The experimental results show that GA K-means may result in better segmentation than other traditional clustering algo- rithms including simple K-means and SOM from the perspective of intraclass inertia. However, this study has some limitations. Though this study suggests intraclass iner tia as a criterion for performance comparison, it is certain that this is not a complete measure for performance comparison of the clustering algorithms. Consequently, the efforts to develop effective measures to compare clustering algorithms should be provided in the future research. In addition, it is also needed to apply GA K-means to other domains in order to validate and prove its usefulness Acknowledgements This work was supported by Korea Research Foundation Grant(KRF-2003-041 B00181). References P.S., Fayyad, U. M: Refining Initial Points for K-means Clustering. Proc. of the ernational Conference on Learning(1998)91-99 Simkin. L: The Market tation Workbook: Target Marketing for Market- 3. Gehrt, K. C, Shim, S: A Shopping Orientation Segmentation of French Consumers: Im- plications for Catalog Marketing. J. of Interactive Marketing 12(4)(1998)34-46414 Kyoung-jae Kim and Hyunchul Ahn 5 Concluding Remarks In this study, we propose new clustering algorithm, GA K-means, and applies it to the market segmentation of electronic commerce. The experimental results show that GA K-means may result in better segmentation than other traditional clustering algo￾rithms including simple K-means and SOM from the perspective of intraclass inertia. However, this study has some limitations. Though this study suggests intraclass iner￾tia as a criterion for performance comparison, it is certain that this is not a complete measure for performance comparison of the clustering algorithms. Consequently, the efforts to develop effective measures to compare clustering algorithms should be provided in the future research. In addition, it is also needed to apply GA K-means to other domains in order to validate and prove its usefulness. Acknowledgements This work was supported by Korea Research Foundation Grant(KRF-2003-041- B00181). References 1. Bradley, P. S., Fayyad, U. M.: Refining Initial Points for K-means Clustering. Proc. of the 15th International Conference on Machine Learning (1998) 91-99 2. Dibb, S., Simkin, L.: The Market Segmentation Workbook: Target Marketing for Market￾ing Managers, Routledge, London (1995) 3. Gehrt, K. C., Shim, S.: A Shopping Orientation Segmentation of French Consumers: Im￾plications for Catalog Marketing. J. of Interactive Marketing 12(4) (1998) 34-46 Table 2. The result of Chi-square analysis K-Means SOM GA K-Means Variables Chi-square Value Asy. Sig. Chi-square Value Asy. Sig. Chi-square Value Asy. Sig. Age 17.867 0.0222 16.447 0.0364 14.598 0.0670 Gender 51.055 0.0000 48.955 0.0000 52.366 0.0000 Income 148.427 0.0000 134.269 0.0000 129.597 0.0000 Education 68.990 0.0002 72.062 0.0001 78.159 0.0000 Occupation 127.347 0.0000 137.115 0.0000 136.304 0.0000 Years of Web use 197.416 0.0000 203.004 0.0000 206.384 0.0000 Frequency of Web search 528.672 0.0000 535.888 0.0000 559.481 0.0000 Number of purchases over the Web 1027.910 0.0000 1003.725 0.0000 1017.909 0.0000 Experience of Web ordering (yes or no) 662.929 0.0000 654.013 0.0000 654.869 0.0000