论文部分内容阅读
目前的局部建模方法在构建样本间相似度的时候仅考虑了输入信息而忽略了输出信息的作用,并且没有考虑样本的权重问题。针对上述问题,提出了局部自适应加权最小二乘支持向量机(Local Adaptive Weight LSSVM,LAW-LSSVM)回归算法。该算法采用同时考虑输入输出信息的相似性判据则来构建更加合理的相似样本集,利用有监督的局部保持映射(Supervised LocalityPreserving Projection,SLPP)算法对样本空间进行有效的降维和搜索最优的相似样本方向,实现了样本权重的在线调整。利用LAW-LSSVM对青霉素发酵过程中的产物浓度进行在线预测,仿真结果表明,包含了输入输出信息的相似度评价准则能够更准确的选择相似样本,较离线LSSVM以及局部LSSVM(LLSSVM)有着更高的预测精度、更好的泛化能力。
The current local modeling methods only consider the input information while ignoring the output information when constructing the similarity between samples, and do not consider the sample weight problem. To solve the above problems, a local adaptive weighted least squares support vector machine (LAW-LSSVM) regression algorithm is proposed. This algorithm uses the similarity criterion which takes into account the input and output information to construct a more reasonable similar sample set, and uses supervised Locality Preserving Projection (SLPP) algorithm to reduce the dimension of the sample space and search the optimal Similar sample direction, online adjustment of sample weight is realized. The method of LAW-LSSVM is used to predict the concentration of penicillin in the fermentation process. The simulation results show that the similarity evaluation criteria including input and output information can select more similar samples more accurately than offline LSSVM and local LSSVM (LLSSVM) The prediction accuracy, better generalization ability.