Density-aware Stratified Sampling for Visualizing Large Volume Geo-Spatial Data

Sampling is a popular approach in big data visualization, however, current sampling approaches don't work well when visualization type is scatter plot, and are even worse in supporting keyword search queries. In this paper, we present an approach of density-aware stratified sampling, it first probing the density of record in different areas of the visualization, then taking the density data to guide the stratified sampling. We conducted an extensively user study to show the efficiency and efficacy of our approach, the experiment shows that our approach can provide very close scatter plots of keyword search queries of a 200 million record dataset within 0.2 second, and the construction time is only 1/4 of an alternative method.