改进K-means算法在多晶硅干法回收氢中研究及应用.docVIP

改进K-means算法在多晶硅干法回收氢中研究及应用 　　 摘要本文针对k-means算法的初始聚类个数K值较难的确定问题提出改进，确定一个最佳聚类个数范围，利用逐个归类法获得初始聚类划分，然后根据函数对初始聚类划分进行估算，降低确定一个较优的K值的难度，并将其应用到多晶硅质量分析中，找出造成干法回收氢不合格的因素，提高产品质量。 　　关键词数据挖掘；聚类；多晶硅；干法回收氢；K-means算法 　　Abstract Aiming at the problem of the initial number k is difficult to determine when using k-means algorithm, an improvement of the algorithm is proposed. To reduce the difficult of determining an optimal value of k, first, determine the range of the best number of cluster and individually classified method to obtain the divided of the initial clustering, then to estimate the initial cluste