论文部分内容阅读
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.
Aiming at the limitations of traditional regressive weighted regression (WDMR) model for predicting new sample data, a WDMR modeling method based on semi-supervised local linear embedding (LLE) algorithm is proposed. Combining with the idea of semi-supervised manifold learning, WDLD model of data-driven semi-supervised LLE algorithm.Then, based on the wheel and rail wear detection data, the tire tread wear prediction experiment is carried out.The results show that compared with the traditional WDMR model, the WDMR model of the semi-supervised LLE algorithm has better Fitting and generalization performance, the prediction accuracy is higher. It is effective to use this model to predict the on-the-spot wheel tread wear.