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

51.71% Mean accuracy

0.5493 Mean F1

0.223733 Mean p-adic loss

62,090 Avg effective params

Total train samples: 53,088
Total test samples: 13,272
Number of tags (input features): 4,875
Number of classes: 657
Avg tree depth: 1121.4
Avg number of nodes: 6,632
Avg number of leaves: 3,317
Effective params formula: nodes × log₂(classes) = 6,632 × 9.36 ≈ 62,090

Cross-validation results

Fold	Accuracy	F1	P-adic loss (mean)	Tree depth	Nodes	Effective params
0	50.66%	0.5412	0.237790	1127	6,499	60,829
1	53.45%	0.5650	0.216004	1118	6,605	61,792
2	51.46%	0.5463	0.218898	1095	6,749	63,169
3	51.47%	0.5463	0.222194	1130	6,615	61,958
4	51.48%	0.5478	0.223780	1137	6,693	62,703

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	ACCESSORIES	0.004096	5/5
2	NEW	0.003277	5/5
3	BIRTHDAY	0.003151	5/5
4	WOMEN	0.003064	5/5
5	KIDS	0.002923	5/5
6	BLACK	0.002870	5/5
7	CLOTHING	0.002674	5/5
8	RETAIL	0.002641	5/5
9	HAT	0.002580	5/5
10	WOMENS	0.002439	5/5
11	DRESS	0.002429	5/5
12	GRANITE	0.002184	1/5
13	PRINTED POCKET SQ	0.002162	3/5
14	POOL	0.002156	1/5
15	OLIO AROMATIZZATO	0.002156	1/5
16	CAR COASTERS	0.002153	3/5
17	PALEO HOT CEREAL	0.002153	3/5
18	INKYBAY_CUSTOM	0.002153	5/5
19	THERMOSTATS	0.002153	5/5
20	RAFCAMERA	0.002153	5/5
21	CHEESE TOOL	0.002153	5/5
22	TOWEL-RAILS	0.002153	5/5
23	NICOTINE	0.002153	5/5
24	ELECTRIC-SKATEBOARDS	0.002153	5/5
25	SIRUP	0.002153	5/5
26	HEALTH AND BEAUTY	0.002153	5/5
27	KEYS	0.002153	5/5
28	WARRANTY-SEATS_2	0.002153	5/5
29	FOOD COLOR	0.002153	5/5
30	PLANT STANDS AND BENCHES	0.002153	5/5
31	SHOP: PELLETS	0.002153	5/5
32	BIKES	0.002153	5/5
33	_RELATED_HOT-CEREAL-LANDING-PAGE	0.002153	2/5
34	CAR ACCESSORIES	0.002152	2/5
35	TYPE_EXTENSION	0.002148	1/5
36	IMT TOOLS	0.002145	4/5
37	PRINTED SILK	0.002138	2/5
38	CATEGORY_BATHMATS	0.002126	5/5
39	SANCTUARY COLLECTION	0.002118	4/5
40	_RELATED_GRANOLA-LANDING-PAGE	0.002089	5/5
41	HOME OFFICE FURNITURE	0.002089	1/5
42	2024	0.002060	2/5
43	WIRE MANAGEMENT	0.002053	5/5
44	BEAN BAG	0.002049	1/5
45	25 MONTHS TO 36 MONTHS	0.002037	3/5
46	BEAN BAG CHAIR	0.002031	2/5
47	__LABEL:NEW	0.002019	5/5
48	HALL50OFF	0.002015	2/5
49	BRUSH	0.001994	5/5
50	TYPE_CORNER	0.001993	5/5