Building an Interpretable Trading Classifier with Python and MQL5

Summary

The article presents a workflow for building a transparent financial machine learning model and connecting it to MetaTrader 5. It contrasts glass-box models with black-box methods, discusses the disagreement that can arise among black-box explanation techniques, and demonstrates feature analysis using an interpretable classifier. The example uses historical daily data, technical features such as ATR and RSI, and a target that indicates whether the next close is higher.

The workflow includes a chronological training and testing split, evaluation of predictions, and exporting the model for use from MQL5 through ONNX. The author emphasizes interpretability as a way to inspect and refine features, while noting that glass-box models can be less flexible and may inherit decision-tree limitations. The article’s example and claims do not establish a profitable strategy; its main contribution is a modeling and deployment process, with predictive performance dependent on the data and validation design.

Key ideas

• Glass-box models expose their decision logic and can make model behavior easier to inspect.
• Different black-box explanation methods may produce conflicting accounts of the same model.
• The example predicts next-day direction using daily market data and technical features.
• Training and testing data should be split chronologically for time series work.
• The article demonstrates exporting an interpretable model for MetaTrader use through ONNX, while noting flexibility limits.

Tags

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