Published in IEEE ICDCS, 2025
We proposed Mosaic, a universal distributed account allocation framework to parallel run allocation algorithm. Upon Mosaic, we further propose Pilot, a lightweight distributed algorithm, to guide the allocation. The efficiency is boosted by 104, while maintaining a similar accuracy.
Recommended citation: IEEE ICDCS 2025.
Published in IEEE ICDE, 2023
We proposed a transaction allocation scheme, TxAllo, for reducing cross-shard transactions and optimizing workload balance in sharded blockchain systems. It investigates a novel graph partition direction to cluster frequent interacted addresses into the same shard and reduces cross-shard transactions from 98% to 17%.
Recommended citation: IEEE ICDE 2023
Published in PAKDD, 2020
We proposed FCP Filter, a Bayesian nonparametric framework for dynamic customer behavior prediction and noise filtering.
Recommended citation: PAKDD 2020.
Published in ICONIP, 2020
We developed an end-to-end model using graph attention networks to predict the next shopping basket by leveraging item correlations.
Recommended citation: ICONIP 2020.