摘要
传统RPCL聚类算法是在随机选取样本的前提下修正权矢量的,没有考虑样本集的空间分布情况。为此,该文提出了一种改进的RPCL聚类算法。该算法引入样本区域密度的概念,根据密度大小按不同的概率选取样本,以修正权矢量。利用文犤1犦中的算例证明了新算法比传统RPCL算法具有更好的聚类速度和精度。最后将算法用于基于RBF神经网络的氧化铝高压溶出过程中溶出率的软测量,仿真结果表明改进的RPCL算法能很好地实现数据样本的聚类,从而提高软测量模型的泛化能力。
In traditional RPCL clustering algorithm,weight vectors are updated under the condition of selecting samples randomly and the space distribution of samples is not considered.In this paper,an improved RPCL algorithm is proposed.In this algorithm,sample region density is introduced.Weight vectors are updated under the condition of selecting samples with different probability according to their density.Example in is used to show that this new algo-rithm has better clustering speed and precision than traditional RPCL algorithm.In the end,this algorithm is used in soft sensing system of Leaching Rate(LR)in the High-Pressure Digestion(HPD)of alumina process based on RBF neural network.Simulation result shows that the improved algorithm can cluster the samples well and improve the generality of the soft sensing model.
出处
《计算机工程与应用》
CSCD
北大核心
2003年第31期30-32,200,共4页
Computer Engineering and Applications
基金
国家863高技术研究发展项目资助(编号:2001AA411040)
