摘要
针对传统流形正则化加权回归(WDMR)模型对新样本数据预测的局限性,提出基于半监督局部线性嵌入(LLE)算法的WDMR建模方法.先结合半监督流形学习的思想,建立了数据驱动的半监督LLE算法的WDMR模型.然后,根据轮轨磨耗检测数据进行了车轮踏面磨耗量的预测实验.结果表明,与传统的WDMR模型比较,半监督LLE算法的WDMR模型具有更好的拟合与泛化性能,预测精度更高,将该模型用于现场车轮踏面磨耗量的预测是有效的.
To deal with the limitation of the traditional weight determination by manifold regu-larization (WDMR) model to predict new sample data,a new model of WDMR is proposed. Firstly, based on semi-supervised local linear embedding (LLE) algorithm theory, the WDMR model of semi-supervised LLE of data driven is established. Then,a numerical simulation study is implemen-ted on the WDMR model of semi-supervised LLE. The simulation results show that the new model has better fitting and generalization performance and higher prediction precision,is effective for pre-dicting the wheel wear volume.
出处
《内蒙古大学学报(自然科学版)》
CAS
CSCD
北大核心
2014年第1期63-67,共5页
Journal of Inner Mongolia University:Natural Science Edition
基金
国家自然科学基金项目(60904049
61263010
61164013
51174091)
江西省青年科学基金(20114BAB211014
20122BAB216026)
江西省教育厅项目(GJJ12316)