Large-scale parallel coordinate data simplifying method based on document embedding model

Abstract

The invention discloses a large-scale parallel coordinate data simplification method based on a document embedding model. The method comprises the steps: carrying out the clustering of data on each attribute axis in a parallel coordinate system, and enabling the same clusters on different coordinate axes in the parallel coordinate system to be regarded as the same word, regarding each data line inserted in the parallel coordinate system as a sentence composed of words, and synthesizing sentences corresponding to all the data lines into a corpus; utilizing a Doc2Vec document embedding model totrain a corpus, wherein each sentence in the corpus is expressed as a high-dimensional vector; projecting the obtained high-dimensional vector to a two-dimensional space and sampling, and finally drawing a data line corresponding to a sampling point in a parallel coordinate system to obtain a simplified parallel coordinate system. According to the method, the continuous semantic association features between the data in the parallel coordinate system are captured through the document embedding model, wherein the features are effectively maintained in the simplification process, so that the simplified parallel coordinate not only can reduce visual redundancy, but also can display the implicit continuous association features in the data to the maximum extent.

Description

technical field [0001] The invention relates to a document expression method based on a document embedding model and a method for simplifying large-scale parallel coordinates, and belongs to the field of information technology. Background technique [0002] Parallel coordinates use the geometric layout of line segments to present multi-dimensional attribute data. Its unique geometric distribution characteristics and superior visual expression performance make it widely used in the exploration and analysis of multi-dimensional data. However, as the scale of multidimensional data increases, a large number of data lines in the parallel coordinate system intersect, overlap and overlap, which seriously interferes with the user's cognition of the original multidimensional data. [0003] Filtering, binding, and sampling are the main methods to solve the visual confusion of large-scale parallel coordinates. Filtering can flexibly select the attribute range of the coordinate axis, t...

AI Technical Summary

Problems solved by technology

However, the traditional sampling algorithm is difficult to maintain the continuous features in the parallel coordinate system, and there are certain limitations.

For example, the sampling lens method proposed by Ellis et al. can alleviate the problem of data overlap in visually confusing areas, but it is difficult to observe the hierarchical category features of the data, especially the contextual features formed between the data passing through the coordinate axes, which are easy to hide and lose visual continuity. sex

Images