T-REX: Transformer-Based Category Sequence Generation for Grocery Basket Recommendation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908870668976128 |
|---|---|
| author | Mokhtari, Soroush Asif, Muhammad Tayyab Zubatiy, Sergiy |
| author_facet | Mokhtari, Soroush Asif, Muhammad Tayyab Zubatiy, Sergiy |
| contents | Online grocery shopping presents unique challenges for sequential recommendations due to repetitive purchase patterns and complex item relationships within the baskets. Unlike traditional e-commerce, grocery recommendations must capture both complementary item associations and temporal dependencies across shopping sessions. To address these challenges in Amazon's online grocery business, we propose T-REX, a novel transformer architecture that generates personalized category-level suggestions by learning both short-term basket dependencies and long-term user preferences. Our approach introduces three key innovations: (1) an efficient sampling strategy utilizing dynamic sequence splitting for sparse shopping patterns, (2) an adaptive positional encoding scheme for temporal patterns, and (3) a category-level modeling approach that reduces dimensionality while maintaining recommendation quality. Although masked language modeling techniques like BERT4Rec excel at capturing item relations, they prove less suitable for next basket generation due to information leakage issues. In contrast, T-REX's causal masking approach better aligns with the sequential nature of basket generation, enabling more accurate next-basket predictions. Experiments on large-scale grocery offline data and online A/B tests show significant improvement over existing systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_06631 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | T-REX: Transformer-Based Category Sequence Generation for Grocery Basket Recommendation Mokhtari, Soroush Asif, Muhammad Tayyab Zubatiy, Sergiy Information Retrieval Machine Learning Online grocery shopping presents unique challenges for sequential recommendations due to repetitive purchase patterns and complex item relationships within the baskets. Unlike traditional e-commerce, grocery recommendations must capture both complementary item associations and temporal dependencies across shopping sessions. To address these challenges in Amazon's online grocery business, we propose T-REX, a novel transformer architecture that generates personalized category-level suggestions by learning both short-term basket dependencies and long-term user preferences. Our approach introduces three key innovations: (1) an efficient sampling strategy utilizing dynamic sequence splitting for sparse shopping patterns, (2) an adaptive positional encoding scheme for temporal patterns, and (3) a category-level modeling approach that reduces dimensionality while maintaining recommendation quality. Although masked language modeling techniques like BERT4Rec excel at capturing item relations, they prove less suitable for next basket generation due to information leakage issues. In contrast, T-REX's causal masking approach better aligns with the sequential nature of basket generation, enabling more accurate next-basket predictions. Experiments on large-scale grocery offline data and online A/B tests show significant improvement over existing systems. |
| title | T-REX: Transformer-Based Category Sequence Generation for Grocery Basket Recommendation |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2603.06631 |