Research On Concept Lattice Granularity Division And Visualization Based On Cloud Model

Since the data in practical applications is often data with diversified sources,large volumes,and diversified formats,the classical formal concept analysis theory that only supports Boolean data will not be applicable.Therefore,how to deal with a variety of data formats,so that the formal concept analysis theory can be used,has become many research concerns.In the diversified format data,continuous-type data is the most frequent in the specified range.At present,the method of dealing with the continuous-type data is mostly Granularity Division,which is the process of dividing the value of formal background attribute into several sub-interval attributes.The core idea of it is to directly divide the value intervals of each attribute item of the interval-type form background by the prior knowledge,or artificially stipulate parameters to divide the grain,and then each fraction is identified and replaced.Finally,the multi-valued attribute single value method is used to process the interval type background into the classic form background.It is also an important research direction to display the background lattice generated by the processed form.As the core data structure in formal theory analysis theory,concept lattice can describe the generalization and specialization relations between concepts.The visualization of the concept lattice Hasse diagram can express the internal relations of this knowledge unit intuitively.Therefore,the concept lattice Hasse layout is clear,beautiful,and can clearly reflect the structural characteristics of the concept lattice is another issue of this study.At present,the traditional concept grid layout algorithm is mainly divided into two-dimensional and three-dimensional layout.This study found that the above research still has the following problems:(1)Although the traditional thinking has solved the problem of the application of the concept lattice theory to interval-continuous data,the human intervention in the sub-granulation process cannot eliminate subjective factors,lack of rationality,and low degree of automation;and each fraction is only data.The collection of samples has no scientific significance.Therefore,it is especially important to study scientifically meaningful,automated methods that are consistent with human cognition.(2)The traditional two-dimensional layout algorithm will suffer from defects such as large horizontal expansion,too many line segments,inflexible grid nodes,and poor human-computer interaction experience in the face of complex concept lattices;although some traditional three-dimensional layout methods solve some,for example,if the horizontal expansion is large and the human-computer interaction experience is poor,there are still problems with the line segment coverage nodes in the Hasse diagram,and the grid structure is not clear and unattractive.In order to solve the above problems,the first part of this paper is to study the method of granularity of the interval format of the concept lattice.Second,the traditional concept lattice of three-dimensional layout method has been improved to form the following two main results.(1)When dealing with the continuous data of massive intervals in the concept lattice,there is a problem that the degree of automation of fractionation is low,subjective factors cannot be excluded,and the scientific problem of fractionation is lacking.A concept lattice fractionation method based on cloud model is proposed..The central idea of this method is to segment the frequency distribution of the original formal background attribute items by using the Gaussian mixture model in statistics to treat each fraction in the interval type granularity as a cloud.The ambiguity parameter in the model is used to determine whether the formed cloud(fraction)is reasonable.In this method,the combination of the concept lattice and the cloud model provides the statistical basis for fractionation,and the fractionation is optimized through the miscibility parameter to avoid manual intervention and improve the degree of automation.This method was applied to the 1980-2017 electronic software sales information experiment.Compared with the fuzzy K-means method in the traditional method in terms of automation degree,fractional mathematics characteristics and rationality measurement parameter DBI,verifying the method of fractionation in this paper.Feasibility and superiority.(2)For the traditional 3D visualization layout algorithm,there is a complex concept lattice on the concept lattice representation,and the graphics are not beautiful enough.A 3D visualization layout algorithm for the concept lattice based on virtual nodes is proposed to improve the visualization of the concept lattice.Aesthetics.The algorithm improves the two-dimensional KK(Kamanda Kawai)layout algorithm by adding virtual nodes and combines it with the layering function of traditional hierarchical algorithms.It effectively solves the problem that the horizontal expansion of the nodes in the layer is too large and the node interconnections are too many to be unclear.The concept lattice structure becomes clear and easy to read.Compared with the traditional three-dimensional layout algorithm,the circular distribution algorithm verifies the superiority of the algorithm to the existing three-dimensional layout algorithm in the visual layout of the concept lattice.