Curve-Shape Clustering (Overlooked Pay)
Unsupervised clustering of raw wireline curve shape (GR, RHOB, NPHI, DT, true resistivity, Vsh, porosity -- NOT Sw or any hydrocarbon flag) across every Phase-4-predicted-Sand interval project-wide, then cross-checked against WellStrat's own hydrocarbon flag and computed Sw for enrichment. The point: a model trained directly on the hydrocarbon flag just re-learns historical picking bias -- clustering with zero reference to that flag can surface curve-shape groups that look like known pay but weren't flagged, i.e. candidate overlooked pay.
What "accuracy" means here: This card does not use the hydrocarbon flag as a training label anywhere -- it's purely a post-hoc cross-check on unsupervised clusters. A cluster with an elevated hydrocarbon-flag rate or unusually low mean Sw relative to the population baseline is a candidate worth manual review, not a confirmed find.
Last run: 2026-08-02 05:20:02 UTC
Curve-Shape Clusters (2D Embedding)
Each point is one Phase-4-Sand interval, positioned by a 2D PCA projection of its own curve-shape features (GR/RHOB/NPHI/DT/Rt/Vsh/PHI -- NOT Sw or the hydrocarbon flag). "By cluster" shows the unsupervised KMeans grouping; "By hydrocarbon flag" recolours the SAME points by WellStrat's own flag, so a cluster that visibly concentrates flagged points is enriched for known pay -- and nearby unflagged points in that same cluster are the overlooked-pay candidates.
| Cluster | Intervals | % HC-flagged | vs. baseline |
|---|
How this result was generated
(project-wide, no formation restriction)"] feat["Mean GR/RHOB/NPHI/DT/log(Rt)/Vsh/PHI
per interval (NOT Sw)"] cluster["KMeans (k chosen by silhouette sweep)"] embed["PCA to 2D (for visualization only)"] check["Cross-check: hydrocarbon flag %,
mean Sw, per cluster vs population baseline"] sand --> feat --> cluster --> embed --> check
Hover a step for a plain-language explanation.