Hierarchical Clustering For Portfolios, Part 2/3

Hierarchical Clustering for Portfolios, Part 2/3 | Dr Krzysztof Ozimek

Title slide: How Hierarchical Clustering Builds a Portfolio, a primer on hierarchical clustering, series 2 of 3 before the workshop

Slide titled What Is Clustering: clustering as unsupervised learning, partitional clustering such as k-means versus hierarchical clustering with nested structure

Slide titled Why Hierarchical: the dendrogram as a binary merge tree, agglomerative versus divisive construction, merge height as dissimilarity

Slide titled Building the Tree, Distances and Linkage: Lopez de Prado and Mantegna distance formulas, and single, complete, group average and Ward linkage rules

Slide titled From Tree to Portfolio, Three Landmark Portfolios: Hierarchical 1/N, Hierarchical Risk Parity (HRP) and Hierarchical Equal Risk Contribution (HERC)

Slide titled A Working Example, Clusters versus Real Industries: table of DJIA and S&P 500 clusters found from price data alone, Mantegna and Stanley (2004)

Slide titled This Is Exactly What the Workshop Builds: hierarchical clustering across the workshop's 12-asset universe before any allocation is built

Packt banner for the hands-on workshop Build and Stress-Test a Multi-Asset Portfolio with Python with Cordell L. Tanny, November 28, 2026, 9:30 AM EDT, with the text I'll be there

A correlation matrix treats every asset pair as an equally plausible substitute for another. That lack of structure is a big part of why naive diversification concentrates risk in ways that aren't obvious until it's too late.

Swipe through a primer on hierarchical clustering: what it is, why it's called "hierarchical," how the dendrogram gets built, and three landmark ways to turn that tree into portfolio weights — plus a real-world check that it actually recovers true industry structure.

This is the second of three short posts I'm sharing ahead of Build and Stress-Test a Multi-Asset Portfolio with Python, a hands-on online workshop with Cordell Tanny, CFA, FRM, FDP on Saturday, November 28, 2026 — where hierarchical clustering is used to replace naive diversification across the workshop's 12-asset universe.

Full details and registration: https://luma.com/e0iytryu?coupon=OZIMEK40&utm_source=expert

Next up in this series: the Johnson SU distribution — and a review after the workshop.

Disclosure: I received a complimentary ticket to this event from Packt Publishing. The views above are my own.