Finding and Classifying Symmetric Fractal Patterns with Machine Learning
Summary
The article proposes detecting scale-varying symmetric patterns in price data through sliding-window correlation. For each starting point, it compares the first half of a window with the reversed, sign-inverted second half across a configured range of even window lengths, retaining the strongest absolute correlation and its associated length. A plotting routine is used to inspect high-correlation candidates. The broader workflow trains a CatBoost classifier using labeled data and describes exporting a resulting model to ONNX for use in MetaTrader 5.
The text presents this as an exploratory forecasting approach and refers to tests on EUR/USD hourly data, including a test on new data, but the supplied excerpt gives no numerical performance results or enough testing detail to assess the evidence. The author cautions that correlation between mirrored historical segments may not capture relationships with future prices and suggests regression as a possible alternative. Pattern selection, label construction, parameter tuning, and out-of-sample testing therefore remain important limits when interpreting predictive claims.
Key ideas
- A sliding window searches for symmetry by correlating one half of a price segment with the reversed, sign-inverted other half.
- The algorithm records the strongest absolute correlation and corresponding window length at each starting point.
- Candidate patterns can be plotted for visual inspection before classifier training.
- A CatBoost classifier is trained from labeled data, with an ONNX export path for MetaTrader 5.
- The article cautions that mirrored-segment correlation may not represent the relationship between past patterns and future prices.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.