Date-Based Grouping for Stock Ranking Models
Summary
This short exchange explains how to assign groups when training a learning-to-rank model on stock observations. The questioner expects each trading date to form a group, so the model ranks stocks against other stocks observed on that date, and asks where that grouping is configured in the stockRanker interface.
The answer says date is used as the grouping key by default, so a separate group setting is not needed for this behavior. This clarifies the intended cross-sectional ranking unit and addresses the concern about validation grouping only implicitly. The note offers no implementation details, documentation about validation-set handling, or empirical comparison of alternative grouping choices; users should not infer more about those mechanics from this brief answer.
Key ideas
- A learning-to-rank model needs to know which stock samples should be ranked against one another.
- For stock ranking, the relevant groups can be formed by trading date.
- The stockRanker described here groups samples by date automatically.
- The exchange does not provide details on validation-set grouping or alternative grouping schemes.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.