Lab
ResearchPublished
One physical feature beat hyper-parameter tuning
In our GSM study, adding Pierce’s tightness factor did more for accuracy than tuning — and SHAP showed the model using it the way textile physics says it should.
Individual machine settings correlate only weakly with fabric weight. Physics says they interact: Pierce's geometry gives a tightness factor, K = √tex / stitch length.
What happened
- Extra Trees with K: 4.10 g/m² RMSE on the test set (R² 0.952).
- Without K: about 4.4 g/m².
- Tuning Extra Trees with randomised search: 4.33 — worse than the defaults with K.
Why explanations matter here
SHAP attributed more than 86% of importance to yarn count (52%), the tightness factor (23%) and stitch length (11%) — the variables a textile engineer would name. When explanations agree with domain physics, the model becomes something a factory can trust, not just a leaderboard entry.