基于初始中心迭代收敛的文本聚类方法  

A Text Clustering Algorithm for Iteration Convergence Based on Initial Centers

作　　者：刘金岭[1] 刘国香[2] 杨凤霞[2] 

机构地区：[1]淮阴工学院计算机工程学院,江苏淮安223003 [2]沧州师范学院计算机系,河北沧州061001

出　　处：《计算机工程与科学》2012年第6期153-157,共5页Computer Engineering & Science

基　　金：河北省科技支撑计划资助项目(10213581);淮安科技计划资助项目(HAG09061);淮阴工学院重点基金资助项目(HGA0907)

摘　　要：利用两三次的K-Means迭代得到初始簇的中心,选择一组具有良好区分度的方向构建IMIC坐标系,在该坐标系下,构造出各坐标轴的重新标度函数用以提高聚类决策的有效性。算法IMIC经过多次迭代,最后收敛到最终解。IMIC算法的时间复杂度与K-Means保持在同一量级上。实验结果表明,IMIC算法有较好的聚类质量。Using K-Means iterative twice K-Means iterative to get an initial cluster center,in the text clustering process, a set of discriminative directions are chosen to construct the IMIC coordinate, each axis is constructed to re-scaling function in order to improve the effectiveness of cluster policy,according to the distribution characteristics of the initial clusters. The IMIC iterative algorithm converges to the final solution. The time complexity of IMIC remains the same as K-Means by using a K-Means-like iteration strategy. The experimental results show that the IMIC algorithm has better clustering quality.

关 键 词：迭代收敛 文本 聚类 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]