摘要
基于邻域的概念,提出一种新的样本筛选方法用于分类问题.该方法在特征空间中根据邻域内的样本类别筛选出具有代表性的训练样本,计算其与测试样本的距离作为样本所属类别的判定依据.在UCI数据集和电力系统负荷预测的应用当中,与SVM和NC两种分类方法进行对比分析,证明该方法能够较好地提高样本识别率并降低时间复杂度.
A new sample filtering method was proposed for classification problems,based on the idea of neighborhood.The representative samples were screened out according to their sorts in neighborhood,then the distance between the representative samples and the test samples was calculated to be the key of classifier.In the experiment,this new algorithm was compared with SVM and NC to validate the classification quality indeed increased and the time complication reduced.
出处
《青岛理工大学学报》
CAS
2010年第5期73-76,共4页
Journal of Qingdao University of Technology
基金
青岛大学青年科研基金项目(2007005)
关键词
空间映射
邻域
样本筛选
电力系统负荷预测
space mapping
neighborhood
sample filtering
electric power system load forecasting
