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

55.21% Mean accuracy

0.5860 Mean F1

0.208079 Mean p-adic loss

40,902 Avg effective params

Total train samples: 36,440
Total test samples: 9,110
Number of tags (input features): 3,664
Number of classes: 489
Avg tree depth: 841.8
Avg number of nodes: 4,578
Avg number of leaves: 2,289
Effective params formula: nodes × log₂(classes) = 4,578 × 8.93 ≈ 40,902

Cross-validation results

Fold	Accuracy	F1	P-adic loss (mean)	Tree depth	Nodes	Effective params
0	56.16%	0.5966	0.195829	846	4,681	41,819
1	53.76%	0.5723	0.213184	840	4,595	41,091
2	55.61%	0.5897	0.210349	831	4,591	41,001
3	54.96%	0.5795	0.204780	846	4,539	40,550
4	55.58%	0.5918	0.216254	846	4,483	40,050

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	BLACK	0.004359	5/5
2	BIRTHDAY	0.004313	5/5
3	KIDS	0.004251	5/5
4	ACCESSORIES	0.003715	5/5
5	CLOTHING	0.003395	5/5
6	HAT	0.003372	5/5
7	DRESS	0.003216	5/5
8	CHRISTMAS	0.002870	5/5
9	SIZE-36R-30W	0.002868	1/5
10	ALL30MLPRODUCTS	0.002859	2/5
11	GRANITE	0.002855	1/5
12	RESPIRATORY	0.002855	1/5
13	MOTORCYCLES	0.002850	1/5
14	PRINTED POCKET SQ	0.002849	3/5
15	SHOP: PELLETS	0.002849	5/5
16	POWER	0.002849	5/5
17	RAFCAMERA	0.002849	5/5
18	HEALTH AND BEAUTY	0.002849	5/5
19	BACKPACK	0.002849	5/5
20	TUBES	0.002849	5/5
21	TOWEL-RAILS	0.002849	5/5
22	CORPORATE	0.002849	5/5
23	GARMIN BIKE MOUNTS	0.002849	5/5
24	KEYS	0.002849	5/5
25	SIRUP	0.002849	5/5
26	PLANT STANDS AND BENCHES	0.002849	5/5
27	WARRANTY-SEATS_2	0.002849	5/5
28	EXTEND	0.002849	5/5
29	THERMOSTATS	0.002849	5/5
30	PRINTED SILK	0.002848	2/5
31	IMT TOOLS	0.002847	4/5
32	YEAST	0.002844	4/5
33	NICOTINE	0.002842	3/5
34	OVER 25	0.002840	1/5
35	INFUSED	0.002831	1/5
36	NEW	0.002781	5/5
37	SIENA	0.002733	2/5
38	EBOOK	0.002727	5/5
39	WHISKEY/WHISKY	0.002722	5/5
40	WIRE MANAGEMENT	0.002716	5/5
41	VINYL CAFE	0.002712	3/5
42	SLIPPERS	0.002711	5/5
43	HALL50OFF	0.002710	2/5
44	BEAN BAG CHAIRS	0.002708	1/5
45	AFRICAN BASKET	0.002704	5/5
46	2024	0.002681	3/5
47	PLATES	0.002680	5/5
48	SUPER CHUNKY BEAN BAG	0.002672	1/5
49	ZERO	0.002662	1/5
50	13 MONTHS TO 24 MONTHS	0.002656	3/5