Working paper

Uganda CIPI Unsupervised Taxonomy

The code, data and full documentation are on GitHub. Last updated 2026-11-20.

The question

Do Uganda's construction input prices really move in the groups that official trade classifications assume?

What's inside
  • Unsupervised machine learning (dynamic time warping and clustering) applied to 18 monthly construction price sub-indices from 2017 to 2026
  • A data-driven taxonomy of price behaviour compared with administrative trade groupings
  • Robustness and stability checks
Why it matters

If prices cluster differently from the official categories, escalation formulas and cost indices built on those categories may misstate risk.

This is a working paper: the analysis is in progress and findings may change. The paper is not yet public; the code and data are on GitHub. Get in touch to discuss the work.