为合理确定批量定制服装的版型数量,运用K-means算法,以4个测量项目(身高、胸围、腰围、领围)为分类变量对347名男性进行聚类分析.分别以国标和非国标号型对初始聚心选择和聚类数的确定进行探讨,并以Calinski-Harabasz(CH)指标、变异系数和相对偏差比较了国标和非国标号型的聚类效果.研究结果表明,运用最大最小距离法确定初始聚心的非国标号型分类结果与国标GB/T 1335.1-2008分类结果对比,在相同CH值时,服装版型数由26减少到18,身高、胸围、领围和腰围相对偏差超过3%的比例分别从5.48%,39.48%,7.49%,60.52%降低到0.58%,8.07%,3.17%,12.97%.测量项目波动性从大到小依次为腰围、领围、胸围和身高.%In order to determine reasonable pattern number of mass customization clothing,the 347-male body data was analyzed based on K-means clustering algorithm with four classified variables such as height (H),bust circumference (BC),waist circumference (WC) and collar circumference (CC).Classification methods of the non-national standard shape and national standard shape to research selection of initial centers and determination of the optimal clusters,and evaluated the cluster in Calinski-Harabasz(CH) index,coefficient of variation and relative deviation.The results show that the clothing pattern number of 26 reduces to 18 and relative deviation(H,BC,CC,WC) is decreased from 5.48%,39.48%,7.49%,60.52% to 0.58%,8.07%,3.17%,12.97%,when the CH index is same and compared to the non-national standard shape that maximum-minimum distance algorithm is adopted to determine the initial centers with national GB/T 1335.1-2008.The volatility of measuring items from big to small is WC,CC,BC and H.
展开▼