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

62.64% Mean accuracy

0.6568 Mean F1

0.231779 Mean p-adic loss

19,494 Avg effective params

Total train samples: 18,320
Total test samples: 4,580
Number of tags (input features): 6,936
Number of classes: 349
Avg tree depth: 624.6
Avg number of nodes: 2,308
Avg number of leaves: 1,154
Effective params formula: nodes × log₂(classes) = 2,308 × 8.45 ≈ 19,494

Cross-validation results

Fold	Accuracy	F1	P-adic loss (mean)	Tree depth	Nodes	Effective params
0	63.12%	0.6549	0.227856	611	2,325	19,639
1	61.31%	0.6478	0.244632	620	2,347	19,806
2	63.51%	0.6754	0.225030	611	2,261	19,089
3	62.64%	0.6568	0.247696	635	2,285	19,330
4	62.60%	0.6491	0.213681	646	2,321	19,606

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	SUPER CHUNKY BEAN BAG	0.003775	1/5
2	TOWEL-RAILS	0.003775	1/5
3	GRANITE	0.003775	1/5
4	PRINTED POCKET SQ	0.003775	1/5
5	INTERCHANGEABLE WATCH BAND	0.003775	1/5
6	MOSS NEEDLE	0.003775	1/5
7	BICYCLES AND MOTORCYCLES	0.003763	1/5
8	EXTEND SHIPPING PROTECTION	0.003763	1/5
9	BEAN BAG	0.003763	1/5
10	CRAFTING SUPPLIES	0.003763	1/5
11	HD TRANSPARENT WIG	0.003763	1/5
12	NISSAN	0.003763	1/5
13	PALEO HOT CEREAL	0.003750	2/5
14	SCRAPBOOKING	0.003747	2/5
15	CAR ACCESSORIES	0.003746	4/5
16	SILICONE WATCH STRAP	0.003744	2/5
17	PIVIT	0.003741	2/5
18	ALL-ACCESSORIES	0.003741	2/5
19	SOUNDBAR	0.003740	3/5
20	SEASONAL	0.003735	4/5
21	KEYS	0.003733	5/5
22	SIRUP	0.003733	5/5
23	ELECTRIC-SKATEBOARDS	0.003733	5/5
24	FOOD COLOR	0.003733	5/5
25	TUBES	0.003733	5/5
26	CHEESE TOOL	0.003733	5/5
27	SPONGES AND HAND PADS	0.003733	5/5
28	BIKES	0.003733	5/5
29	CATEGORY_CANVASPRINTS	0.003733	5/5
30	RAFCAMERA	0.003733	5/5
31	CIDER	0.003733	5/5
32	PLANT STANDS AND BENCHES	0.003733	5/5
33	FINIAL	0.003733	5/5
34	JIGS & RIGS	0.003729	2/5
35	STYLE-SWIM	0.003728	3/5
36	LIGHTWEIGHT	0.003726	4/5
37	MOTORCYCLE RAMP	0.003726	4/5
38	TUMBLER SUPPLIES	0.003725	1/5
39	TV SPEAKER	0.003725	1/5
40	TRAILER JIGS	0.003725	1/5
41	MALE	0.003725	1/5
42	BATHWARE	0.003725	1/5
43	KNITTED CHAIR	0.003725	1/5
44	ZOULOU	0.003722	2/5
45	PRINTED SILK	0.003722	4/5
46	IMT TOOLS	0.003722	4/5
47	_RELATED_HOT-CEREAL-LANDING-PAGE	0.003721	3/5
48	PLATES	0.003720	5/5
49	HALL50OFF	0.003720	5/5
50	AMBER BEADS	0.003719	1/5