Evaluation of Sampling Methods for Scatterplots

Given a scatterplot with tens of thousands points or even more, natural question is which sampling method should be used to create small but “good” for better abstraction. We present the results user study that investigates influence different strategies on multi-class scatterplots. The main goal this understand capability methods in preserving density, outliers, and overall shape scatterplot. To end, we comprehensively review literature select seven typical as well eight representative datasets. then design four experiments performance maintaining: 1) region density; 2) class 3) outliers; 4) results. show that: random preferred blue noise have comparable three outlier biased density based sampling, recursive subdivision perform best keeping outperforms others maintaining