High-Frequency Equity Factors and GRU-Based Weekly Signals

Summary

This weekly report surveys Chinese equity-selection factors built from intraday trading behavior. Examples include return skewness, downside-volatility share, buying interest and large-order net buying after the open, late-session trading share, average outflow size, large-order-driven price moves, and an improved reversal measure. It also lists four deep-learning factor configurations combining gated recurrent units with neural networks. Conventional factors are described as orthogonalized, and the report outlines long-short portfolios formed from the top and bottom deciles, with monthly rebalancing for the traditional factors and weekly rebalancing for the deep-learning variants.

The text says the source report summarizes weekly, February, and 2022 long-short returns, long-side excess returns, and monthly win rates. However, the figures and tables containing those results are absent here, so no factor’s performance or the headline 2022 return can be independently assessed from this document. Factor construction is largely deferred to referenced research. The stated risks are factor decay and liquidity, and the material provides no detailed implementation or transaction-cost analysis.

Key ideas

• The report surveys intraday-derived equity factors, including trade-flow, volatility, and return-distribution measures.
• It describes long-short portfolios using the top and bottom deciles of factor rankings.
• Traditional factors are shown with monthly turnover, while the listed GRU-neural-network variants use weekly turnover.
• The source references performance tables, but those results are not present in the supplied text.
• Factor decay and liquidity are identified as risks.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.