Selecting Specific Lookback Windows for Batch Factor Generation
Summary
The document explains how to generate moving-average features in BigQuant when only selected lookback windows are wanted. It contrasts a range-based list comprehension, which the platform accepts, with a list of chosen values, which it says is unsupported in that form. The proposed workarounds are to enter each feature separately in the input feature list or generate the full range of features and then filter the resulting DataFrame with a custom module.
The discussion is a brief platform-support answer rather than a general treatment of factor design. It gives no example of implementing the DataFrame filtering step, and it reports no tests, performance results, or guidance on choosing lookback periods. The advice is therefore specific to the described BigQuant feature-input workflow and may depend on the platform's current behavior.
Key ideas
- BigQuant is described as rejecting a generated feature list built from an explicit set of lookback values.
- A range-based list comprehension is reported to work for generating moving-average features.
- One workaround is to enter each desired feature separately in the input feature list.
- Another workaround is to generate a broader feature DataFrame and filter its columns in a custom module.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.