Discussing the article: "Feature Engineering for ML (Part 2): Implementing Fixed-Width Fractional Differentiation in MQL5"

[MetaQuotes]

Check out the new article: Feature Engineering for ML (Part 2): Implementing Fixed-Width Fractional Differentiation in MQL5.

This article delivers a production-grade MQL5 implementation of fixed-width fractional differentiation for live MetaTrader 5 feeds. We introduce a header-only CFFDEngine that precomputes weights without a fixed cap, performs O(width) per-bar updates, and avoids per-tick allocations. The FFD.mq5 indicator supports all ENUM_APPLIED_PRICE types and prev_calculated optimization. Validation scripts confirm numerical equivalence with the standard Python frac diff_ffd pipeline.

Part 1 developed the theory and Python implementation of fractional differentiation using the fixed-width window (FFD) method from Chapter 5 of AFML. Three properties of the FFD method make it ideal for live deployment: the weight vector is precomputed once, each observation depends on a bounded lookback window, and the computation is a single dot product. This article translates those properties into a production-grade MQL5 engine that runs efficiently on live MetaTrader 5 data feeds.

The implementation has two components: CFFDEngine (a header-only .mqh class for weight generation and dot-product computation) and FFD.mq5 (a custom indicator that wraps CFFDEngine and draws the fractionally differentiated series). The indicator supports MetaTrader's prev_calculated optimization and recomputes only what has changed since the last call.

The design goal throughout is: zero per-tick allocation, O(width) per new bar, and O(1) initialization amortized. On a typical instrument with d = 0.4 and τ = 10⁻⁵, the threshold-determined window width is 1457 bars. At τ = 10⁻⁴ it falls to 281 bars. The dot product over that window completes in microseconds on modern hardware.

Author: Patrick Murimi Njoroge