Quarterly Indicator Selection and MLP Forecasting for the XLF ETF

Summary

The article extends a quarterly indicator-ranking pipeline into a predictive model for the XLF ETF. It uses FrAMA, Parabolic SAR, Alligator lines, and TRIX as aligned features, then trains a multilayer perceptron to classify future returns or predict them as a regression target. The labels are based on forward returns over a fixed horizon, with bullish, bearish, and neutral classes separated by a configurable threshold. The intended workflow exports the trained model to ONNX for inference in an MQL5 Expert Advisor.

The method emphasizes preprocessing and feature discipline: segment observations by quarter, normalize indicator readings, align multi-line outputs, and avoid leaking future information into earlier rows. The article describes recent-data validation and model deployment, but the supplied text does not provide detailed numerical performance results or enough information to assess robustness. Its claims about regime generalization should therefore be treated as a research objective; performance may depend on threshold choices, validation design, and the XLF’s market history.

Key ideas

• The pipeline uses quarterly indicator rankings to inform features for an MLP forecasting model.
• FrAMA, Parabolic SAR, Alligator components, and TRIX provide complementary trend and momentum inputs.
• Forward returns are converted into bullish, bearish, or neutral labels using a configurable threshold.
• The model is intended for ONNX export and use inside an MQL5 Expert Advisor.
• Quarter-aware validation and careful feature alignment are important, but the text gives limited quantitative evidence of out-of-sample performance.

Tags

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