ProSeCo: Visual analysis of class separation measures and dataset characteristics

Class separation is an important concept in machine learning and visual analytics. We address the analysis of class measures for both high-dimensional data its corresponding projections into 2D through dimensionality reduction (DR) methods. Although a plethora have been proposed, it difficult to compare between multiple datasets with different characteristics, measures, DR present ProSeCo, interactive visualization approach support comparison up 20 4 methods, respect any 7 dataset characteristics: size, dimensions, counts, size variability, skewness, outlieriness, real-world vs. synthetically generated data. ProSeCo supports (1) comparing across (2) dimensionally-reduced (3) methods (4) partitioning characteristic, (5) partitions selected characteristic (6) inspecting individual detail. demonstrate utility two usage scenarios, using [1] posted at https://osf.io/epcf9/.