Using Monoid Actions and Feature Importance for Trailing-Stop Models

Summary

The article connects monoid actions, which let operations transform values in a separate set, to a machine-learning system that forecasts changes in price-bar range for adjusting trailing stops. It discusses feature-importance methods, including tree-based importance, permutation importance, and SHAP, as ways to examine the roles of inputs such as lookback period, timeframe, applied price, indicator, and trade decision. The proposed use is to identify influential inputs and expand a monoid’s choices through additional action values, then assess whether forecasts change.

Illustrative rankings and performance drops are presented, followed by a suggested train/test workflow with categorical encoding, gradient boosting, and evaluation metrics. These figures are examples in the text, not independently supported empirical findings. The article also notes that implementing gradient boosting in MQL5 can be difficult, and the available material is incomplete. Feature importance alone does not establish causality or prove that expanded inputs improve out-of-sample trading results.

Key ideas

• Monoid actions allow a model’s operation to map from a monoid into a broader set of transformed values.
• Feature-importance methods can help prioritize inputs for further analysis.
• The article proposes expanding lookback or indicator choices when those features appear influential.
• Training and testing on separate data is suggested to assess forecast performance.
• The reported rankings are illustrative and do not establish trading profitability.

Tags

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