Combining Volatility and Volume Features with Quantum Encoding for FX Forecasting

Summary

This article describes an experimental machine-learning system for forecasting movement across eight currency pairs on a 15-minute timeframe. It combines four groups of inputs: 3D-bar features derived from price, time, volume, and volatility; a quantum circuit encoder; conventional technical indicators; and CatBoost, with an optional language-model interpretation layer. The 3D-bar features use cyclical hour encoding, returns and their changes, volume changes, and rolling volatility measures to represent market behavior in normalized form.

The article proposes a “yellow cluster” signal when price and volume volatility both exceed their 70th percentiles, interpreting the condition as a possible reversal zone. It reports tests on over 400,000 EURUSD bars from 2022 to 2024, and claims improved accuracy, win rate, and return after adding the 3D-bar module. These results are presented without enough detail here to establish robustness or out-of-sample reliability. The method also has potential leakage concerns: percentile thresholds use the full series, and the reversal score uses a centered window. Parameters are tuned for EURUSD M15, processing is slow, and periodic retraining is said to be necessary.

Key ideas

• The system combines 3D-bar features, quantum encoding, technical indicators, and CatBoost for FX direction forecasts.
• The 3D-bar representation encodes time cyclically and derives price, volume, and volatility changes from OHLCV data.
• The yellow-cluster rule flags simultaneous high price and volume volatility as a possible reversal condition.
• The reported evaluation covers EURUSD bars from 2022 to 2024, but the article does not establish broad out-of-sample validity.
• Full-sample percentile thresholds and a centered reversal window may introduce information from future observations.

Tags

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