【作者】 任俊利；

【导师】 洪文学；

【作者基本信息】 燕山大学 ， 生物医学工程， 2010， 硕士

【摘要】 本文针对海量数据处理上存在的问题和难点进行分析,提出了多源海量数据特征融合的多层次递阶结构,设计了一种处理多源海量数据的模型。此模型主要用来处理高维数据递阶、分层降维和分类决策的问题。第一章主要介绍了海量数据处理和数据融合的背景,强调海量数据处理和数据融合在生物医学、水力部门、金融等方面的重要性。本章基于生物医学领域,阐述了医学数据的庞大和无序,急需要有专业的数据处理技术来处理海量的数据。本文的主要研究内容是提出一种海量数据的处理方法,能够从海量的数据中挖掘出隐含的信息。第二章主要描述了海量数据的特性和对海量数据预处理的方法。海量数据一般是在超维空间中表示,而超维空间与日常中的低维空间有着明显的区别。低维空间的数据主要分布在空间的中部,而高维空间却恰恰相反,数据主要分布在边缘。所以降维的主要依据就是将数据从超维到低维的转化。数据的类型是多源的,为了将不同数据进行融合和比较,需要将不同的数据进行统一化,归一化。此过程中数据的预处理(如格式转换、归一化、降噪、平滑、降维等)就显得非常重要了。第三章介绍了基于分层递阶图的降维与可视化信息融合方法,本章重点介绍分层递阶图的模型,海量数据降维和可视化信息融合等,可视化信息融合的方法。第四章在前几章的理论基础上,提出了一种基于多源海量数据分层递阶图表示模型。并且将此模型运用到脑电数据的可视化处理上,得出仿真结果,说明本文提出的模型的可行性。将此模型的仿真结果与文献中已有模型的仿真结果进行对比,体现本文提出的模型的优越性。更多还原

【Abstract】 Massive data processing in domestic and international problems and difficulties of a full analysis, this mass of data for the medicine, the main mass of data of high dimensional data hierarchy, hierarchical dimension reduction problem. Firstly it explains the high dimensional data reduction the background, importance and urgency. How to make high-dimensional data dimension reduction algorithm using a more appropriate medical treatment and analysis of massive data, this problem has been a hot topic in today’s medical profession, but there are many domestic and foreign experts and scholars have done in this area out a welcome contribution. This paper is to study the issue of huge amounts of data preprocessing, the use of massive data preprocessing is appropriate, is the key to the success of biomarker identification. In this study, also made reference to many domestic and foreign research methods, pericoin the application of cluster analysis combined with genetic algorithms. Lilien principal component analysis and linear discriminant method. Morris the use of wavelet transform and peak detection algorithm. Yu using high-throughput mass spectrometry data algorithm development.This article focuses on the geometric properties of massive data, given the type of mass data, and dimensionality reduction and data features of the mathematical description of the concept. Secondly, the paper introduces a hierarchical graph-based dimensionality reduction and visual information fusion method to introduce a hierarchical graph models, information visualization, massive data reduction and fusion of visual information visualization information fusion method, and various methods were simple comparison shows the advantages and disadvantages of each method.更多还原