FactorGCL combines a hypergraph neural network with a cross-temporal contrastive loss to mine hidden factors for stock return prediction, reporting SOTA IC/ICIR on China A-shares.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
q-fin.ST 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
FactorGCL: A Hypergraph-Based Factor Model with Temporal Residual Contrastive Learning for Stock Returns Prediction
FactorGCL combines a hypergraph neural network with a cross-temporal contrastive loss to mine hidden factors for stock return prediction, reporting SOTA IC/ICIR on China A-shares.