摘要
该文用RBF神经网络建立了转炉提钒冷却剂预报模型。RBF网络的中心的选取采用了可以在线学习的最近邻聚类算法。为了进一步优化网络中心 ,提出了基于密度排名的最近邻聚类算法。该算法聚类前先将样本按其在样本空间的密度进行了排序 ,聚类过程始于样本空间最密集处。实践证明 ,该算法应用于提钒冷却剂预报模型的建立是合理的 ,可行的。
This paper surveys a kind of Radial Basis Function (RBF) algorithm to establish the vanadium extraction prediction model. To optimize the location of RBF centers, we introduced an improved nearest-neighbor clustering algorithm based on density ranking method. This algorithm ranks the patterns by their density, and then the nearest-neighbor clustering algorithm is applied to the patterns. This improved algorithm starts the clustering process from the densest pattern space, and therefore it effectively eliminates the bad impact of the noisy data. The improved algorithm proved effective in the vanadium extraction intelligent decision system modeling.
出处
《计算机仿真》
CSCD
2003年第11期55-57,共3页
Computer Simulation
基金
国家教育部博士点基金项目 (980 6117)
重庆市基础研究项目 (73 69)
