CASE improves next-basket repurchase recommendation by modeling item cadences as calendar-time signals with multi-scale convolutions and induced set attention, delivering up to 8.6% precision and 9.9% recall lifts at top-5 in production.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
CASE: Cadence-Aware Set Encoding for Large-Scale Next Basket Repurchase Recommendation
CASE improves next-basket repurchase recommendation by modeling item cadences as calendar-time signals with multi-scale convolutions and induced set attention, delivering up to 8.6% precision and 9.9% recall lifts at top-5 in production.