T-REX: Transformer-Based Category Sequence Generation for Grocery Basket Recommendation

TL;DR

T-REX is a transformer-based model designed for grocery basket recommendation, effectively capturing both short-term and long-term shopping patterns through innovative sampling, encoding, and category-level modeling, leading to improved prediction accuracy.

Contribution

The paper introduces T-REX, a novel transformer architecture with unique sampling, encoding, and category-level modeling tailored for grocery basket recommendation.

Findings

Significant improvement over existing recommendation systems.

Effective modeling of both short-term and long-term shopping patterns.

Successful deployment in large-scale offline and online environments.

Abstract

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.