Using UMAP to Combine Multiple Indicator Periods in MQL5

Summary

The article addresses the challenge of choosing a technical indicator’s lookback period: short periods can react to noise, while long periods may lag. It proposes feeding readings from many periods into a dimension reduction method, focusing on UMAP to represent the combined data in fewer features for a market forecasting model. The example replaces an earlier RSI implementation with Williams Percent Range and describes building shared MQL5 classes to manage indicator buffers, differences between readings, and model use through ONNX.

The discussion also introduces object oriented design, particularly inheritance, as a way to share functionality across single buffer indicators. It presents conceptual claims about UMAP’s ability to preserve nonlinear structure and contrasts it with linear approaches, but the supplied text is incomplete and does not include enough of the analysis or trading results to assess predictive performance. The method’s usefulness therefore depends on careful data preparation, validation, and avoiding overfitting; using every available period does not itself establish an edge.

Key ideas

• Short indicator periods can capture noise, while long periods can make signals late.
• The article proposes combining readings from multiple indicator periods with UMAP dimensionality reduction.
• Its MQL5 example uses Williams Percent Range values and describes using ONNX models.
• A shared parent class can centralize functionality for indicators that use one data buffer.
• The supplied text does not provide enough performance evidence to judge the forecasting approach.

Tags

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