Spanning Trees as Approximation of Data Structures

The connections in a graph generate structure that is independent of coordinate system. This visual metaphor allows creating more flexible representation data than two-dimensional scatterplot. In this article, we present STAD (Simplified Topological Abstraction Data), parameter-free dimensionality reduction method projects high-dimensional into graph. generates an abstract by giving each point location which preserves the approximate distances original space. built upon Minimum Spanning Tree (MST) to new edges are added until correlation between from and dataset maximized. Additionally, supports inclusion additional functions focus exploration allow analysis perspectives, emphasizing traits otherwise would remain hidden. We demonstrate effectiveness our applying it two real-world datasets: traffic density Barcelona temporal measurements air quality Castile León Spain.