Training a Neural Network in NeuroPro and Exporting It to MetaTrader 5

Summary

This tutorial shows how to build a basic neural-network trading Expert Advisor by connecting MetaTrader 5 with NeuroPro, a multilayer network tool. It exports EURUSD hourly closing prices to CSV, converts the data to DBF for NeuroPro, trains and tests a network, then carries its learned weights into an MQL5 implementation. The example uses the preceding 24 bars to forecast the direction over the next hour and applies a minimum forecast threshold to filter trades whose expected move is small relative to spread.

The article reports a growing equity curve in a historical MetaTester run, but offers no evidence that performance persists outside the training period. It explicitly cautions that in-sample profit does not establish live profitability and that robust systems require understanding both neural-network behavior and trading. The tutorial is also tied to older software and Windows versions the author tested; compatibility with later versions is not established.

Key ideas

• The example exports historical prices from MetaTrader 5 and converts them to a format NeuroPro can read.
• A multilayer sigmoid network is trained and tested on separate data arrays to help assess overfitting.
• The tutorial converts learned network weights into MQL5 so the Expert Advisor can make predictions directly.
• The example predicts the next hour's direction from recent hourly prices and filters forecasts below a threshold.
• A favorable historical equity curve does not demonstrate performance on unseen data or a live account.

Tags

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