Time Series Analysis and Forecasting

DTMiner optimizes temporal motif mining by coordinating multiple matching tasks around shared data access patterns, achieving 1.14×-11.98× performance improvements.

STORM combines dual vector quantized autoencoders to extract temporal and spatial stock features, creating diverse factor embeddings for improved asset pricing and portfolio management.

Temporal-clique subgraph pattern matching via integrated topological and temporal selectivity optimization with specialized join operators and cardinality estimation.