Mining Price-Volume Factors with Gene Expression Programming

Summary

This report summary describes using gene expression programming, a heuristic inspired by biological evolution, to discover price-volume factors from short-horizon data. Because many candidate factors had weakly monotonic group returns or no excess return in the long portfolio, the authors screened candidates using a combination of information coefficient information ratio, long-side excess returns, and monotonicity across return groups. The summary says the selected factors showed more pronounced and orderly selection behavior.

The mined weekly factors were combined with traditional fundamental factors to form long-only and CSI 500 enhanced portfolios. The supplied summary reports annualized returns or excess returns, information ratios, and yearly outperformance claims for weekly and daily rebalancing variants, and says results improved relative to fundamental-factor portfolios without the mined factors. These are reported historical results; the excerpt provides no underlying sample period, detailed construction rules, transaction-cost treatment, or independent validation. It also flags systemic market risk and the possibility that factor effectiveness may change, so the reported performance should not be assumed to persist.

Key ideas

• Gene expression programming is used as a heuristic search method for price-volume factors.
• Factor screening combines ICIR, long-side excess returns, and monotonicity across return groups.
• The mined weekly factors are combined with fundamental factors in long-only and CSI 500 enhanced portfolios.
• The summary reports historical performance improvements but omits detailed validation and implementation assumptions.
• The report identifies market-wide risk and changing factor effectiveness as caveats.

Tags

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