A daily unsupervised machine-learning read of the US Equity market: 4,785 stocks grouped into 10 consensus clusters (KMeans + Gaussian-mixture + hierarchical, over robust-scaled PCA features) for 2026-07-15. Descriptive, not predictive — there is no buy or sell signal. Research, not investment advice.
high margin · low volatility (3m) · large cap — 1,700 names; mostly Financial Services; drivers: profit margin (+0.46), rv 63 (−0.40), log market cap (+0.39)
high downside vol · high volatility (1y) · high leverage — 783 names; mostly Healthcare; drivers: vol of vol 63 (+0.68), distance from high 252 (−0.68), downside vol 63 (+0.60)
falling earnings · rising (3m) · expensive (low E/P) — 593 names; mostly Healthcare; drivers: ni growth (−1.11), ret 63d (+0.99), ret 20d (+0.76)
expensive (low E/P) · low cash yield · high leverage — 351 names; mostly Healthcare; drivers: val ep z (−4.36), val cfop z (−3.50), vol of vol 63 (+1.33)
high turnover · high growth · high beta — 249 names; mostly Technology; drivers: turnover to mcap (+2.04), revenue growth (+1.61), beta 252 d (+1.40)
rising (3m) · strong 1y momentum · high volatility (3m) — 240 names; mostly Healthcare; drivers: ret 63d (+2.48), vol of vol 63 (+2.25), ret 126d (+2.05)
high cash yield · cheap (high B/P) · cheap (high E/P) — 213 names; mostly Financial Services; drivers: val cfop z (+4.18), val bp z (+4.07), val ep z (+3.94)
high downside vol · high volatility (3m) · high volatility (1y) — 182 names; mostly Healthcare; drivers: vol of vol 63 (+4.02), downside vol 63 (+3.47), rv 63 (+3.22)
expensive (low E/P) · high leverage · high cash yield — 169 names; mostly Consumer Cyclical; drivers: val ep z (−3.54), leverage debt to mcap (+3.50), val cfop z (+2.85)
Machine-readable data (free, read-only JSON)
The full map, per-ticker cluster assignments with confidence and anomaly scores, PCA structure, and within-cluster cointegrated relative-value pairs are published as open JSON for automated and AI-analyst consumption:
/api/ml/latest — clusters, names, sector mix, representative tickers, top features
/api/ml/pairs — within-cluster cointegrated pairs (Engle-Granger, OU half-life, Kalman z-score; US only)
Descriptive market-structure research only. Unsupervised clustering finds structure, not direction; a tight cluster or an anomaly is a starting point for research, never a trade signal.