Hierarchical clustering

In search of data-driven constellations

Concept, graphic design


Politecnico di Milano

InDesign, Illustrator, Python, Plotly, SciPy

Michele Bruno, Maria Celeste Casolino, Luca Draisci, Virginia Leccisotti, Barbara Roncalli, Sara Zanardi

Since time immemorial, mankind has looked at the sky, gazing at heavenly bodies and connecting the closest ones to draw figures in the sky. This is how constellations originated. What would constellations look like if, instead of humans, it was an algorithm that searched for patterns in the celestial vault?

This poster applies information design principle to explain what is hierarchical clustering, a famous statistical method. We decided to use build our own constellations using this statistics and then we dreamed about what mythological being they could represent.
How to better visualize clusterization than by just… clustering? To explain this statistical method we decided to apply it to the most clustered items in the history of mankind: stars.
Taking in consideration the brightest stars in the sky, we played with some Python libraries to create a customized algorithm and visualize our data-driven constellations.

Repository on GitHub ↗︎

All rights reserved
© 2022 Alice Bocchio