Decision Tree Classifier

Model overview

Decision tree classifier for predicting taxonomy IDs from product tags. Unlike parameter-constrained models, this classifier uses ALL available tags. The tree structure provides natural interpretability - each decision node tests whether a specific tag is present, leading to a taxonomy prediction at the leaves.

5 CV folds

57.17% Mean accuracy

0.6001 Mean F1

0.219522 Mean p-adic loss

35,430 Avg effective params

Total train samples: 30,760
Total test samples: 7,690
Number of tags (input features): 5,904
Number of classes: 551
Avg tree depth: 894.0
Avg number of nodes: 3,892
Avg number of leaves: 1,946
Effective params formula: nodes × log₂(classes) = 3,892 × 9.11 ≈ 35,430

Cross-validation results

Fold	Accuracy	F1	P-adic loss (mean)	Tree depth	Nodes	Effective params
0	56.34%	0.5967	0.214507	880	3,857	35,121
1	58.93%	0.6168	0.213862	895	3,899	35,463
2	57.94%	0.6044	0.201248	892	3,861	35,158
3	55.69%	0.5825	0.255987	896	3,881	35,350
4	56.94%	0.6000	0.212008	907	3,961	36,058

Feature importances (Top 50)

Feature importances are averaged across all 5 folds. The "Folds" column shows how many folds used this feature.

Rank	Tag	Avg Importance	Folds
1	NEW	0.003648	5/5
2	BIRTHDAY	0.003511	5/5
3	ACCESSORIES	0.003105	5/5
4	NEW ARRIVALS	0.003104	5/5
5	DRESS	0.002575	5/5
6	HAWRYCH RETAIL	0.002507	1/5
7	FULL	0.002507	1/5
8	JIGS & RIGS	0.002507	1/5
9	BICYCLES AND MOTORCYCLES	0.002494	1/5
10	INTERCHANGEABLE WATCH STRAP	0.002494	1/5
11	GRANITE	0.002494	1/5
12	HOME THEATER SPEAKERS	0.002492	3/5
13	PALEO HOT CEREAL	0.002492	2/5
14	PRINTED SILK	0.002492	2/5
15	CAR ACCESSORIES	0.002491	2/5
16	INTERCHANGEABLE WATCH	0.002491	2/5
17	HD TRANSPARENT WIG	0.002490	1/5
18	US-SIZE-9-5	0.002490	1/5
19	SOUNDBAR	0.002490	1/5
20	FINISHING OILS	0.002490	1/5
21	INTERCHANGEABLE WATCH BAND	0.002490	1/5
22	AMBER	0.002490	1/5
23	TUBES	0.002488	5/5
24	PLANT STANDS AND BENCHES	0.002488	5/5
25	HEALTH AND BEAUTY	0.002488	5/5
26	TOWEL-RAILS	0.002488	5/5
27	SIDE CHAIR 2000	0.002488	5/5
28	ELECTRIC-SKATEBOARDS	0.002488	5/5
29	KEYS	0.002488	5/5
30	FINIAL	0.002488	5/5
31	RAFCAMERA	0.002488	5/5
32	CHEESE TOOL	0.002488	5/5
33	SIRUP	0.002488	5/5
34	FOOD COLOR	0.002488	5/5
35	__LABEL:NEW	0.002488	5/5
36	CIDER	0.002488	5/5
37	TIRE SEALANT	0.002488	5/5
38	WARRANTY-SEATS_2	0.002488	5/5
39	SPONGES AND HAND PADS	0.002488	5/5
40	LIGHTWEIGHT	0.002488	4/5
41	US-SIZE-10-5	0.002488	4/5
42	IMT TOOLS	0.002487	4/5
43	BIKES	0.002487	4/5
44	MOTORCYCLE RAMP	0.002487	4/5
45	CAR COASTERS	0.002486	3/5
46	PIVIT	0.002486	3/5
47	_RELATED_HOT-CEREAL-LANDING-PAGE	0.002486	3/5
48	PRINTED POCKET SQ	0.002486	3/5
49	MOUNTAIN BIKES	0.002481	3/5
50	ZOULOU	0.002480	3/5