K.j. Kim, H. Ahn/Expert Systems with Applications 34(2008)1200-1209 the time of this research. Thus, it is impossible to apply CF order to predict the items which are likely to be purchased in our target shopping mall because the buying behavior of by users. However, CBR also has a limitation that is simi the customers cannot be patternized. Consequently, we lar to CF-that is the requirement of high computational applied a simple case-based reasoning(CBR)model in complexity as the user base increases. To resolve the prob lem, clustering methods have been applied as a preprocess- ing tool for CBR(Kim Han, 2001). In this study, we Table 2 Intraclass inertia of each clustering method used our proposed clustering algorithm-GA K-means Clustering method as a preprocessing tool for CBR. Fig 3 shows the system GA K-means architecture of the recommendation system using GA K- Intraclass inertia 2.1467 2.1288 means and CBr The collected data in this case included 3298 customers It contained users' demographic variables and the items Table 3 that they had purchased. We collected totally 42 indepen t-values of th es f-test dent variables including demographic and other personal GA K-means information. Table I presents detailed information on the 5.5344 features In order to determine the most appropriate cluste Significant at the l% level lgorithm for this data set, we also applied simple K-me Table 4 Table 5 The result of Chi-square analysis The result of anova Code K-means SOM GA K-mean SOM value AGE93.4440.00020988000002170180.000 96.3680.0001348.8310.00010.235 HEIGHT ADDI 13.4680009270.957000092780.055 WEIGHT ADD2 14833200005164660000144920006OBJ 57.055000012.2480.01672.6850.000 BMI OCCUR OCCUR 450.737 834180.000167.6360.000 WAIST 0.000312450.000 OCCU2 l079419 372.9070.0001458.2390.000 OC cCU 3.6630.00820.0050.000 OC cCU 13+20 02.1150.000727.3880.000 214.6960.000184.0980.000 MARRIE GAKI LOSS2 36 216.4060.000351.1340.000 0.000478.5910.000 0.000 0.000390.2100.000 0.000 0.000724.5110.000 0.000 380.218 0.000473.0030.000 0.0001189.1620.000 LOSS6 3393980.000 0.0003777120.000 LOSS 5460720.0006586960.0006119350.000 220.42400003894920.000447.0960.000 LOSS 522.7610.0001145.6260.0001065.2250.000 PURO 15552980.0001666.5330.0001665.7390.000 PURI 141.5330.00089.0950.000134.9720.000 PURe 1772.3690.0001643.8520.0002162.9270.000 144.1860.000157.6860.000156.7710.000 88 96.1150.000169.6920.00099.7730.000 21.6300.00019.8250.00128.1300.000 20.4240.00030.5890.00037.7420.000 EOI 1588750.000105.2900.000110.9690.000 237.7830000113.7550.000971060.000 54.6770.00038.5900.000479420000 0.000116.3980.000144.5750.000 E05 90.4720.000726610.000 0.00015.3720.004 17.5440.002 0.005 0.00030.3690.00018.0630001 27.8190.000 8.0890.0012134 Fig. 4. Clustering result of Ga K-meansthe time of this research. Thus, it is impossible to apply CF in our target shopping mall because the buying behavior of the customers cannot be patternized. Consequently, we applied a simple case-based reasoning (CBR) model in order to predict the items which are likely to be purchased by users. However, CBR also has a limitation that is simi￾lar to CF – that is the requirement of high computational complexity as the user base increases. To resolve the prob￾lem, clustering methods have been applied as a preprocess￾ing tool for CBR (Kim & Han, 2001). In this study, we used our proposed clustering algorithm – GA K-means – as a preprocessing tool for CBR. Fig. 3 shows the system architecture of the recommendation system using GA K￾means and CBR. The collected data in this case included 3298 customers. It contained users’ demographic variables and the items that they had purchased. We collected totally 42 indepen￾dent variables including demographic and other personal information. Table 1 presents detailed information on the features. In order to determine the most appropriate clustering algorithm for this data set, we also applied simple K-means, Table 2 Intraclass inertia of each clustering method Clustering method K-means SOM GA K-means Intraclass inertia 2.1467 2.1340 2.1288 Table 3 t-values of the paired-samples t-test SOM GA K-means K-means 5.534a 8.233a SOM 3.588a a Significant at the 1% level. Table 4 The result of Chi-square analysis Code K-means SOM GA K-means Chi￾square value Asy. Sig. Chi￾square value Asy. Sig. Chi￾square value Asy. Sig. ADD0 96.368 0.000 1348.831 0.000 10.235 0.037 ADD1 13.468 0.009 270.957 0.000 9.278 0.055 ADD2 148.332 0.000 516.466 0.000 14.492 0.006 OCCU0 57.055 0.000 12.248 0.016 72.685 0.000 OCCU1 450.737 0.000 83.418 0.000 167.636 0.000 OCCU2 1079.419 0.000 372.907 0.000 1458.239 0.000 OCCU3 1.937 0.747 13.663 0.008 20.005 0.000 OCCU4 603.448 0.000 202.115 0.000 727.388 0.000 SEX 167.612 0.000 214.696 0.000 184.098 0.000 MARRIED 661.327 0.000 216.406 0.000 351.134 0.000 LOSS1 757.606 0.000 478.591 0.000 593.393 0.000 LOSS2 445.300 0.000 390.210 0.000 416.854 0.000 LOSS3 1075.998 0.000 724.511 0.000 740.688 0.000 LOSS4 380.218 0.000 443.821 0.000 473.003 0.000 LOSS5 696.827 0.000 1305.146 0.000 1189.162 0.000 LOSS6 339.398 0.000 414.516 0.000 377.712 0.000 LOSS7 646.072 0.000 658.696 0.000 611.935 0.000 LOSS8 220.424 0.000 389.492 0.000 447.096 0.000 LOSS9 522.761 0.000 1145.626 0.000 1065.225 0.000 PUR0 1555.298 0.000 1666.533 0.000 1665.739 0.000 PUR1 141.533 0.000 89.095 0.000 134.972 0.000 PUR2 1772.369 0.000 1643.852 0.000 2162.927 0.000 D1 144.186 0.000 157.686 0.000 156.771 0.000 D2 96.115 0.000 169.692 0.000 99.773 0.000 D3 21.630 0.000 19.825 0.001 28.130 0.000 D4 20.424 0.000 30.589 0.000 37.742 0.000 E01 158.875 0.000 105.290 0.000 110.969 0.000 E02 237.783 0.000 113.755 0.000 97.106 0.000 E03 54.677 0.000 38.590 0.000 47.942 0.000 E04 153.452 0.000 116.398 0.000 144.575 0.000 E05 86.949 0.000 90.472 0.000 72.661 0.000 E06 27.896 0.000 15.372 0.004 17.544 0.002 E07 14.877 0.005 16.113 0.003 26.981 0.000 E08 51.238 0.000 25.048 0.000 32.696 0.000 E09 38.677 0.000 30.369 0.000 18.063 0.001 E10 27.819 0.000 18.089 0.001 21.357 0.000 Table 5 The result of ANOVA Code K-means SOM GA K-means F- value Asy. Sig. F- value Asy. Sig. F- value Asy. Sig. AGE 93.444 0.000 209.880 0.000 217.018 0.000 HEIGHT 31.631 0.000 22.773 0.000 23.624 0.000 WEIGHT 23.181 0.000 26.752 0.000 29.990 0.000 OBJ 19.671 0.000 21.144 0.000 23.581 0.000 BMI 37.419 0.000 41.808 0.000 54.352 0.000 WAIST 25.189 0.000 31.245 0.000 40.390 0.000 Fig. 4. Clustering result of GA K-means. 1206 K.-j. Kim, H. Ahn / Expert Systems with Applications 34 (2008) 1200–1209