Parameter Constrained Logistic Regression

Predicting Shopify taxonomy from product tags using parameter constrained logistic regression

← Back to index

Model summary

This parameter constrained logistic regression model predicts taxonomy IDs from product tags. Trained at: 2026-03-20T05:30:47+11:00

5CV folds

7.94%Mean test accuracy

493Taxonomy classes

Aggregate statistics

Total train samples: 36,440
Total test samples: 9,110
Mean F1 (weighted): 0.0844
Mean p-adic loss: 0.66497540

Cross-validation fold results

Average p-adic loss across all folds: 0.66497540

Fold	Test Accuracy	F1 Score	P-adic Loss (mean)	Details
0	8.30%	0.0866	0.78003279	View details →
1	7.25%	0.0801	0.74281185	View details →
2	7.72%	0.0858	0.54223690	View details →
3	7.70%	0.0814	0.54853147	View details →
4	8.71%	0.0882	0.71126399	View details →

Largest taxonomy classes

Taxonomy ID	Path	Samples	Share
gid://shopify/TaxonomyCategory/bt	4	94	1.03%
gid://shopify/TaxonomyCategory/lb	15	82	0.90%
gid://shopify/TaxonomyCategory/bu	5	30	0.33%
gid://shopify/TaxonomyCategory/na	25	17	0.19%
gid://shopify/TaxonomyCategory/sg	23	13	0.14%
gid://shopify/TaxonomyCategory/os	18	12	0.13%
gid://shopify/TaxonomyCategory/gc	11	11	0.12%
gid://shopify/TaxonomyCategory/hg	14	8	0.09%
gid://shopify/TaxonomyCategory/ma	16	7	0.08%
gid://shopify/TaxonomyCategory/fb	9	6	0.07%
gid://shopify/TaxonomyCategory/me	17	6	0.07%
gid://shopify/TaxonomyCategory/pa	19	6	0.07%
gid://shopify/TaxonomyCategory/el	8	5	0.05%
gid://shopify/TaxonomyCategory/bi	6	4	0.04%

Tags with strongest signal

Tag	Taxonomy ID	Path	Weight	Max \|weight\|
FPM	gid://shopify/TaxonomyCategory/sg	23	3.9876	3.9876