Spatiotemporal Signals in Residuals: A Two-Stage Boosting Graph Neural Network for Stock Return Prediction
Liu and co-authors (2026) study machine learning return prediction in equities using a feedforward neural network. The abstract mentions a Sharpe ratio but not a benchmark, an out-of-sample test, transaction costs, multiple testing or the size of the universe.
“Under the reported evaluation protocol on the CSI 1000 and CSI 2000 universes, STB-GCRN improves IC, ICIR, gross annualized long–short return, and Sharpe ratio over LightGBM.”
AssetEquities
StrategyMachine learningfactor investing
TechniqueFeedforward networkgraph neural network
FrequencyQuarterly
Also stated: Sharpe ratio
Missing from abstract: benchmark, out-of-sample test, costs, multiple testing, universe
Institutional-grade audit of this strategy.
£5 — Request the audit- Audit of the paper claims accounting for costs, luck, crises, originality, geography, etc.
- Reproduction code integratable into Bloomberg, IBKR, etc.
- Bias-free datasets for backtesting included
- Delivered in 2 working days
The paper's neighbourhoodMachine learning return prediction by year
The family is counted across the whole universe. Asset, method and authors are counted inside the family.