基于最小熵的高维电力数据可视化特征映射方法
|
摘要
针对传统特征映射方法存在映射时间长、高维数据转换率低等问题,提出基于最小熵的高维电力数据可视化特征映射方法。对高维电力数据进行空间模拟,从数据预处理、转换、离散化分析和特征分类方面入手,完成对高维电力数据可视化特征分类。建立电力数据类的散布矩阵,根据矩阵计算高维电力数据的特征相对值和判别值,完成数据特征提取。基于上述特征分类和特征提取结果,利用熵对高维电力数据各类的可分性进行描述,选取出熵最小的数据特征,定义数据的熵并将熵当作数据类别的可分性判据,利用电力数据的总体熵实现高维数据到低维数据的映射。实验结果表明,所提方法的特征数据分类准确度较高,且平均高维数据转换率为78%左右,映射耗时短,远远优于传统方法,验证了所提方法的优越性能。
Aiming at the problems of traditional feature mapping methods,such as long mapping time and low conversion rate of high-dimensional data,a visual feature mapping method based on minimum entropy for high-dimensional power data is proposed. The spatial simulation of high-dimensional electric power data is carried out. The visual feature classification of high-dimensional electric power data is completed from the aspects of data preprocessing,transformation,discrete analysis and feature classification. The scatter matrix of electric power data class is established and the characteristic relative value and discriminant value of high dimensional power data are calculated according to the matrix to complete the data feature extraction. Based on the above features classification and feature extraction results,entropy is used to describe the separability of all kinds of high dimensional power data,and the lowest entropy is selected. Data features define the entropy of the data and regard entropy as the criterion of separability of data categories. The mapping of high-dimensional data to low-dimensional data is realized by using the total entropy of electric power data. The experimental results show that the proposed method has high classification accuracy,and the average conversion rate of high-dimensional data is about 78%. Themapping time is short and the proposed method is much better than the traditional method. The proposed method is proved to be superior to the traditional method.
Aiming at the problems of traditional feature mapping methods,such as long mapping time and low conversion rate of high-dimensional data,a visual feature mapping method based on minimum entropy for high-dimensional power data is proposed. The spatial simulation of high-dimensional electric power data is carried out. The visual feature classification of high-dimensional electric power data is completed from the aspects of data preprocessing,transformation,discrete analysis and feature classification. The scatter matrix of electric power data class is established and the characteristic relative value and discriminant value of high dimensional power data are calculated according to the matrix to complete the data feature extraction. Based on the above features classification and feature extraction results,entropy is used to describe the separability of all kinds of high dimensional power data,and the lowest entropy is selected. Data features define the entropy of the data and regard entropy as the criterion of separability of data categories. The mapping of high-dimensional data to low-dimensional data is realized by using the total entropy of electric power data. The experimental results show that the proposed method has high classification accuracy,and the average conversion rate of high-dimensional data is about 78%. Themapping time is short and the proposed method is much better than the traditional method. The proposed method is proved to be superior to the traditional method.
引文
[1]胡启国,汪文珺.核最小均方算法的特征映射和参数选择[J].南方农机,2017,48(13):136-137.
[2]徐绕山,王爽,孙正兴.大规模艺术图像的视觉特征计算组织与可视化[J].科技通报,2017,33(12):194-200.
[3]冉冉,胡楠,刘鹏宇,等.基于实时历史数据库的电力系统CIM模型研究[J].电子设计工程,2017,25(10):158-161.
[4]左开伟,刘建平,程馨莹.基于降维Householder变换的多任务运动想象脑电信号特征提取研究[J].科学技术与工程,2016,16(20):206-211.
[5]孙劲光,丁胜锋.鲁棒性监督等距特征映射方法[J].中国矿业大学学报,2017,46(4):932-938.
[6]韩璞,张婷,董泽,等.最小熵分布估计算法系统辨识及应用[J].中国电机工程学报,2017,37(21):6363-6372.
[7]李立. 65°三角翼亚音速复杂流场计算和数据可视化[J].力学与实践,2017,39(1):18-24.
[8]陈海辉,周向东,施伯乐.基于稀疏正则化的高维数据可视化分析技术[J].计算机应用与软件,2017,34(6):22-26.
[2]徐绕山,王爽,孙正兴.大规模艺术图像的视觉特征计算组织与可视化[J].科技通报,2017,33(12):194-200.
[3]冉冉,胡楠,刘鹏宇,等.基于实时历史数据库的电力系统CIM模型研究[J].电子设计工程,2017,25(10):158-161.
[4]左开伟,刘建平,程馨莹.基于降维Householder变换的多任务运动想象脑电信号特征提取研究[J].科学技术与工程,2016,16(20):206-211.
[5]孙劲光,丁胜锋.鲁棒性监督等距特征映射方法[J].中国矿业大学学报,2017,46(4):932-938.
[6]韩璞,张婷,董泽,等.最小熵分布估计算法系统辨识及应用[J].中国电机工程学报,2017,37(21):6363-6372.
[7]李立. 65°三角翼亚音速复杂流场计算和数据可视化[J].力学与实践,2017,39(1):18-24.
[8]陈海辉,周向东,施伯乐.基于稀疏正则化的高维数据可视化分析技术[J].计算机应用与软件,2017,34(6):22-26.