摘要
构造了CD4与HIV变化率的评价函数,运用在线聚类法对评价函数值进行聚类,确定了各个标准的阈值,通过建立分类器,可进一步预测治疗效果。在求解模型时采用leader-follower聚类算法,该算法针对在线聚类的特点,只对与新到样本最相似的一个聚类中心进行调整,与该样本无关的其他类的性质得到了保留。实验数据表明,与传统聚类算法相比,该算法达到了可塑性与稳定性的平衡。
The evaluation function based on the change rate of CD4 and HIV was constructed. The on - line cluster method was applied to make cluster analysis of the values of the evaluation function. The threshold of each criterion was got and the effect for further therapeutics was predicted The Leader - follower algorithm ,aiming at the characteristics of on - line cluster analysis, only adjusts the center of one class similar to the newly available sample and reserves features of other classes. The data of the experiment indicate that compared with the traditional cluster methods, this algorithm can achieve the balance of plasticity and stability.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2008年第2期236-239,共4页
Journal of Wuhan University of Technology：Information & Management Engineering
基金
科技部技术创新基金资助项目(02C26214200218)
