Using Association Rules to Generate Forex Trading Signals

Summary

The article presents a workflow for mining associations in hourly data for four major currency pairs. It exports historical bars, calculates returns and technical features such as moving averages, RSI, MACD, volatility, Bollinger bands, and candle patterns, then discretizes observations into categories for rule discovery. Association rules connect combinations of market conditions with later price moves; support, confidence, and lift are used to evaluate candidate relationships before converting them into trading signals.

The author reports pair-level results, including USDJPY with a 65% profitable-trade rate and a 1.6 reward-to-risk ratio, and GBPUSD with 58% and 1.4. Rules with lift above 2.0 and confidence above 0.8 are said to perform best across pairs. These are historical claims without enough information in the excerpt about validation design, transaction costs, or out-of-sample testing to establish reliable live performance. The author identifies dynamic rule settings, macroeconomic information, and adaptation to market phases as areas for further work.

Key ideas

• The workflow combines MQL5 data export with Python feature engineering and rule mining.
• Price and indicator data are discretized into categories before association rules are generated.
• Support, confidence, and lift help assess candidate links between conditions and subsequent moves.
• The article reports different historical outcomes across currency pairs but limited validation detail.
• The author proposes dynamic rules, macroeconomic inputs, and market-phase adaptation as improvements.

Tags

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