ml_proj

Machine Learning with Matlab 2018. Final project on Timeseries Prediction with LSTM / RNN.

0. Project directory structure:

ml_proj/ +-- data % Contains EURUSD dataset +-- scripts % MATLAB Scripts | +-- data % data preprocessing scripts | +-- measure % RMSE and MAPE implementation | +-- arimaMain.m % Run ARIMA | +-- varmMain.m % Run VAR | +-- lstmUnivariateMain.m % Run LSTM univariate | +-- lstmMultivariateMain.m % Run LSTM multivariate | +-- config.m % Configuration +-- reports % Report (report.pdf) +-- saved_models % Trained models

1. Installation In order to run the project, the following Matlab Toolboxes must be installed: 1. Statistics and Machine Learning Toolbox 2. Econometrics Toolbox 3. Deep Learning Toolbox

2. Running the Project

   • Add ml_proj and its subfolders into path
   • Modify the configuration if necessary. By default, it only verifies the saved models in full dataset. Set cfg.execMode to "train" to train the model again.
   • Run the respective .m file