Double-target sampling visual analysis method for large-scale social media data

Abstract

The invention discloses a double-target sampling visual analysis method for large-scale social media data, and the method comprises the steps: carrying out the semantic correlation training of the large-scale geographic space social media data through a Doc2vec model, converting the large-scale geographic space social media data into a high-dimensional vector, and then mapping the high-dimensionalvector to a low-dimensional space for visual presentation; defining a double-target sampling visual analysis method, an adaptive blue noise sampling algorithm being used for sampling original socialmedia data in a semantic space, and effectively reserving semantic correlation in the original social media data; in order to keep the spatial distribution of the social media data at the same time, quantifying the spatial distribution of the sample data by using an adaptive quadtree method, and replacing semantic spatial sampling points in an iterative manner, so as to achieve a dual purpose of optimizing semantic and spatial distribution at the same time; and further designing a semantic round interaction means in a map window to support a user to effectively evaluate the effectiveness of adouble-target sampling algorithm, so that semantic and spatial distribution characteristics are analyzed and mined in an exploratory manner for large-scale geographic space social media data.

Description

technical field [0001] The invention belongs to the technical fields of geospatial data analysis, graphics and visualization, and in particular relates to a visual analysis method for dual-target sampling of large-scale social media data. Background technique [0002] With the rapid growth of social media data, it has become possible to explore the important information contained in it through visual analysis technology. However, the visual mapping elements of social media data overlap each other in the map view, largely disturbing the visual perception of semantic features and their geospatial distribution. Therefore, it is of great significance to reduce the overlapping coverage of large-scale social media data and enhance the visual expression of local semantic features. [0003] However, although existing technologies use visual analysis methods to explore large-scale social media data, the increase in the amount of data will cause the data presentation results to easil...

Images