README.md

Training your model (Running the Code)

This code is intended to run in order:

Forex Data Download

I have used this software to download the Forex one minue data: Quant Data Manager.

Download a Data Set

The model is expecting GBPUSD data set called gbpusd.csv in ./LSTM-FX-Train-Test/data, however, you may use any Forex pair as long as the header (first line) is this:

Date,High,Low

And the rest of the lines have data following this format:

2010-01-01 00:01,1.61674,1.61670

Setting Up the Variables (Parameters and Hyperparameters)

Edit common_variables.py and set the batch_size, window_size, validation_size, test_size. Make sure any other Jupyter Notebook is closed to ensure they pick the latest common_variables.py.

Every time you modify the common_variables.py you will have to start again from this step.

Preparing and Splitting the Data Set

Open prep_and_split.ipynb with Jupyter Notebook and execute it.

Training Your Model

Open train_model.ipynb and modify the epoch number. Your model is generated in ./LSTM-FX-Train-Test/models and your scaler in ./LSTM-FX-Train-Test/scalers.

Testing Your Model

Open test_model.ipynb model and run it. The will test the model for a single time unit prediction.

Multiple Predictions

Open multi_pred_model.ipynb and configure:

• pred_interval: This is to try to predict multiple times separated by this interval.
• pred_size: This is the length of the prediction, in other words, how many units (minutes in this case) do you want to predict into the future.