Building Concept-Group Factors from Industry and Stock Data

Summary

The document explains how to construct factors grouped by Chinese stock market concepts when a built-in grouping expression does not recognize the concept field. For standard industry groupings, it gives a group-mean example using a level-two industry classification and returns. For concept groupings, it recommends deriving the field from available data tables and creating a custom calculation.

The illustrated workflow reads industry membership and daily bar data over a date range, joins them by date and instrument, then groups closing prices by date and concept to calculate a mean. This demonstrates how concept-level aggregates can be formed from security-level observations. The note is a brief implementation answer rather than a factor-research guide: it does not discuss handling stocks assigned to multiple concepts, missing membership data, weighting choices, normalization, or whether the resulting aggregate predicts returns. Those design and validation choices remain necessary when turning the example into a usable factor.

Key ideas

• Industry factors can be grouped with a mean aggregation over an industry classification.
• Concept fields may need to be derived from underlying industry and bar data.
• Joining membership and daily price data by date and instrument enables concept-level aggregation.
• The example computes an equal-weighted mean closing price for each date and concept.
• The note does not address overlapping concept memberships or predictive validation.

Tags

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