摘要
针对接地网腐蚀速率的小样本及参数建模问题,建立了一种非参数集群分类预测模型。采用接地网腐蚀速率等级分类策略降低了预测模型的复杂度;采取自助法(Bootstrap)产生自举子集,避免小样本问题;结合非参数—KNN分类法和Adaboost法,对所有自举子集建立多个弱分类器,并集群成强分类器。实验结果表明,与KNN分类法相比较,非参数集群算法得到的腐蚀速率等级和实测分类可以较好吻合,该模型适用于接地网腐蚀速率的预测。
A non-parametric cluster classification prediction model was established aiming at small sample and parametric modeling of earth mat corrosion rate. First of all, a earth mat corrosion rate level classification strategy was used to reduce the complexity of the prediction model; secondly, a self-help method (Bootstrap) was utilized to produce Bootstrap subsets, avoiding small sample problem; finally, multiple weak classifiers were generated for all bootstrap subsets with nonparametrie - KNN classification and the Adaboost method and were clustered into a strong classifier. The experimental result shows that compared with the KNN classification method, corrosion rate levels resulting from the nonparametric cluster algorithm can be matched with the actual results better, and this model is suitable for the earth mat corrosion rate prediction.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第12期4362-4367,共6页
Computer Engineering and Design
关键词
接地网腐蚀
小样本
多分类
自助法
非参数法
集群学习
grounding grid corrosion rate
small sample
multiple classification
bootstrap
nonparametric method
cluster learning