基于多视角缺失补全算法的数据挖掘研究
    点此下载全文
引用本文:黄 裕.基于多视角缺失补全算法的数据挖掘研究[J].计算技术与自动化,2018,(2):67-72
摘要点击次数: 23
全文下载次数: 
作者单位
黄 裕 (广东生态工程职业学院广东 广州 510520) 
中文摘要:针对数字化信息复杂度带来的海量多视角数据问题,并考虑到在大量的多视角数据的获取过程中,由于收集的难度、高额成本或设备故障等情况,往往会导致多视角数据出现视角缺失。提出了一种基于核回归的多视角数据缺失补全方法,采用离线核回归模型学习和在线多视角缺失数据补全构建了算法框架,通过引入高斯核核函数的方式,建立视角间的非线性回归模型,结合训练数据的线性组合来表示回归系数的最优解,以完成挖掘多视角数据间的互补相关性,有效实现缺失视角的补全。最后通过模拟三类数据集来验证基于多视角缺失补全算法的性能。
中文关键词:机器学习  多视角数据  视角缺失  核回归  核函数
 
Research on Mining Multi View Deletion Completion Algorithm Based on Data
Abstract:Aiming at massive problems of multi-view data brought by the complexity of digital information, and taking the situation of large number of multi-angle data acquisition into account. Some factors which include the difficulty of collection, high cost or equipment failure often lead to multi-angle data miss. In this paper, a new method of multi-view data missing completion based on kernel regression is proposed. The algorithm framework is constructed by using offline kernel regression model and complement online multi-angle missing data. By introducing Gaussian kernel function, the paper established a nonlinear regression model between perspectives. Combined with the linear combination of training data to represent the optimal solution of the regression coefficient to complete the complementary correlation between the mining multi-angle data, and effectively realize the missing view. Finally, the performance of the algorithm based on multi-view missing completion is verified by simulating three kinds of data sets.
keywords:machine learning  multi view data  lack of view  kernel regression  kernel function
查看全文   查看/发表评论   下载pdf阅读器