This project implements a Forex price prediction model using Long Short-Term Memory (LSTM) neural networks. It consists of two main scripts: train.py for training the model and eval.py for real-time price predictions.

• Python 3.7+
• MetaTrader5
• PyTorch
• pandas
• numpy
• scikit-learn
• matplotlib
• joblib

You can install the required packages using:

pip install MetaTrader5 torch pandas numpy scikit-learn matplotlib joblib

The train.py script fetches historical Forex data, preprocesses it, trains an LSTM model, and saves the trained model and scaler.

• Fetches historical data from MetaTrader5
• Adds technical indicators (EMA, MACD, RSI, Bollinger Bands, ATR, CCI)
• Implements an LSTM model with dropout for better generalization
• Uses early stopping to prevent overfitting
• Saves the best model and scaler for later use

python train.py

You can modify the following parameters in the script:

• SYMBOL: The Forex pair to train on (default: "EURUSD")
• TIMEFRAME: The timeframe for data (default: mt5.TIMEFRAME_M1)
• NUM_CANDLES: Number of candles to fetch (default: 10000)
• SEQ_LENGTH: Sequence length for LSTM input (default: 60)
• BATCH_SIZE: Batch size for training (default: 16)
• EPOCHS: Maximum number of training epochs (default: 100)
• LEARNING_RATE: Learning rate for the optimizer (default: 0.001)

The eval.py script loads the trained model and makes real-time predictions on live Forex data.

• Fetches live data from MetaTrader5
• Uses the trained model to make price predictions
• Calculates prediction accuracy and direction accuracy
• Provides real-time plotting of actual vs predicted prices

python eval.py

You can modify the following parameters in the script:

• SYMBOL: The Forex pair to evaluate (default: "EURUSD")
• TIMEFRAME: The timeframe for data (default: mt5.TIMEFRAME_M1)
• SEQ_LENGTH: Sequence length for LSTM input (default: 30)
• PREDICTION_INTERVAL: Time between predictions in seconds (default: 60)

Both scripts provide logging information and generate plots:

• train.py creates:

  • training_validation_loss.png: Plot of training and validation loss
  • actual_vs_predicted_test.png: Plot of actual vs predicted prices for the test set
  • actual_vs_predicted_all.png: Plot of actual vs predicted prices for the entire dataset

• eval.py provides:

  • Real-time console output with current price, predicted price, and accuracy metrics
  • Live-updating plot of actual vs predicted prices

• Ensure that MetaTrader5 is installed and running on your system.
• The model's performance can vary depending on market conditions and the specific Forex pair being analyzed.
• This project is for educational purposes only and should not be used for actual trading without thorough testing and risk management strategies.

MIT License