A Project that involves training, testing, evaluating and forecasting time series forex data from MetaTrader using Sockets to connect with Python. An RNN Model of various architecture like LSTM, Bidirectional and GRU can be created. Supports CUDA Computation.
MQL based Expert Advisors are attached to charts of trading platforms to make RNN based forecasts. All machine learning parameters can be tuned from the input interface of the EA. The model can be trained and exported as a serialized .hdf5 file. Tests are run simultaneously with initial train and the data is split based on Testing Part (in %) parameter. The model can be evaluated on any number of metrics as provided in the input screen of the EA. The model can be loaded to retrain or forecast any number of future bars.
The various parameter that can be opted for tuning the ML model are -
- Long Short Term Memory
- Gated Recurrent Units
- Bidirectional LSTM
- Bidirectional GRU
- Mean Sqaure Error
- R^2 Score
- RMSProp (Recommended)
- Stochastic Gradient Descent (Momentum can be specified)
- Adam
- Adagrad
Specify amount of change to the model during each step of this search process, or the step size
For faster computations, GPU can be used. CUDA Support is required. CuDNN implemented RNN layers are used if GPU is opted
To run the forecaster, run the socketserver.py and wait for socket to be created. Now, attach the EA to the MT4/MT5 platform and specify the parameters for building the model. The predicted values are displayed on the same chart window.
python socketserver.py
To install tensorflow-gpu use python package manager
pip install tensorflow-gpu
If tensorflow-gpu is present, run the following commands to check
import tensorflow as tf
tf.test.is_gpu_available()- Python 3.6 or higher (dev versions not recommended, use stable releases)
- Tensorflow 1.14 or higher
- Keras
For GPU Support
- tensorflow-gpu
- CUDA 9.0 or higher (10.0 recommended)
- cuDNN Library v7.4 or higher
