| Data is the manifestation and carrier of information,and is the original figure and facts.Data has been growing and accumulating in great speed along with the fast development of information technology,especially the rapid development of techniques including information acquisition,internet,internet of things,social networks.A large amount of data is wide spreading in all trades and professions.There is tremendous value in big data.Those data begin to bringing benefit to human beings,which turns into significant wealth of information society indicating the age of big data is coming.In 2015,China proposed the Action Outline for Promoting Mass Data Development which designs the overall development for our country`s theories and techniques of big data from high level.In 2016,our nation puts forward the “Thirteen-Five” National Scientific and Technological Innovation Plan which explicitly highlights the big data project will be initiated and the mass data industry clusters with global competitive advantages will be established.Thus,the research and application of big data theory and technology have been advanced to the important strategic position of our nation.Big data has the characteristics of large volume,rapidly updating,multiple modal,difficultly identifying,great value and low density.Big data contains great value.Therefore,the main purpose of big data analysis is to discover the hidden knowledge behind the data.Visual analysis is a significant method for big data analysis.Visual analysis of big data aims to use the computer automatic analysis and simultaneously exploit adequately the cognitive advantages of human to the visual information,combining the strengths of human and computer with the help of human-computer interactive analysis and interactive technology to help people more directly and efficiently insight the information,knowledge and wisdom behind big data.The ten challenges of visual analysis in the future mainly focus on the core topics of visual analysis: cognitive,visualization,deep integration of human-computer interaction.With the development of cognitive science,how to create a visual representation of big data for matching mentalimage in order to analysts see through big data will be the biggest challenge.Granular computing has developed based on human cognitive law.It is the subject basing on problem solving of information granules and its structure,information processing theory,method,technique and tool and is also a new computing paradigm in current field of intelligent information processing.It combines many theoretical research productions including the rough set,fuzzy set and artificial intelligence and so on.Characteristics of big data exploiting and granular computing paradigm have a high degree of consistency.Granular computing will be used as an important method for big data analysis and exploitation.Multidimensional mixed data as a common type of data in large data,it has important scientific value and practical significance.Under the guidance of cognitive science and on the basis of the objective regulations of human cognition,this thesis intends to build a visual grain structure of multidimensional mixed data under the framework of granular computing,explore the key issues on the visual analysis of partial-order structure,try to construct the basic theoretical framework of visual analysis of partial-order structure,research the visual analysis of multidimensional mixed data in view of the partial-order structure.Concrete researches are as follows:First,construct the hierarchical structure of the single attribute grain and its visualization method based on the purpose of exploring the implicit information in the relationship among the attributes.The goal of multidimensional data analysis is to explore the distribution and patterns of multidimensional data and to reveal the hidden relationships among different dimensions.It is significant to dig the hidden relationships among the attributes in the information table.It can provide the single attribute constraint of object class and upper and lower approximate description,describes the generation principle of the granular structure,and design automatic generation algorithm in computers and the visualization method.The meaning of the elements such as the node,path and group structure in this type of granular structure is has been explored,and the calculation method of attribute importance based on attribute grain hierarchy is presented.Secondly,according to the objective law of human cognition,attribute partial-orderstructure diagram with a multi-level and multi-granularity particle structure is put forward,and its formation principle is described.People's knowledge of the objective things is based on the characteristics of things,which are also called attributes.The cognition of human beings to different things is characterized by the cognitive law from macro to micro,from integrity to specificity.And it follows the cognitive mechanism of "looking for the same from the difference,finding the difference from the same,the same close to the same,the difference far away from the difference".From two cognitive perspectives of "global view" and "local view",a generation method for granules is introduced and two kinds of attribute partial-order structure diagram are constructed in order to explore the upper and lower approximation of granules and its physical meaning.Besides,the corresponding structure of the computer automatic generation algorithm and visualization method is designed.Thirdly,the method of the simplification and completion for object model is proposed.The human memory of things doesn't include all features and details of the things and attributes which are universal and most specific are often used.Under the premise of guarantee of mode separability,the simplification for the object model can reduce the attributes used to describe the model.And based on feature attribute and attribute calculation,a method of the model simplification is proposed.Through discussing the completion of object model in information table and using the binary relation between the model and the attributes,the model is constructed on the basis of iterative calculation of formal concept.Finally,a preliminary system that has a core function on visual analysis of the multidimensional and mixed data is designed.The thesis demonstrates the application of this particle structure and algorithm on real data through the combination of different field data. |
PDF Full Text Request