Final Project for GA
Forecast Volatility for EURUSD (most tradable currency pair) for the following timeframes (time in UTC):
- 1day timeframe
- 1hour timeframe (to be continued)
Time Serires - Regression problem. The input shape will be 1 time step.
- DATA + ARIMA.ipynb
- Present Date and demonstrate ARIMA model
- ARIMA Disadvatages (needs: complete data, linear relationship, univariate data)
- Results
- Univariate LSTM and Tuning Parameters.ipynb
- Univariate LTSM model demonstration
- Tune parameters (Epoch, batch sizes, neurons, add 1 more layer)
- Compare RMSE
- Multivariate LSTM.ipynb
- Use the optimised univariate LSTM model and add more features
- Compare results with the optimised Univariate LSTM model
Features:
- EURUSD past volatility
- EURUSD volume
- USDJPY past volatility
- USDJPY volume
Economic Figures
-
EUIR: Interest Rate EUR
-
EUCONF: ECB conference
-
USIR: Interest Rate USD
-
FEDCONF: FED conference
-
USGDP: GDP USD
-
EUGDP: GDP EUR
-
USCPI: CPI USD
-
EUCPI: CPI EUR
-
NFP: NFP
-
EUUR: Unemployment rate EUR
-
USR: retail Sales
-
USCC: US consumer confidence
-
USADP: US ADP Employment Change
-
USISM: US ISM Manufactiring PMI
-
GPMI: German manufacturing PMI
-
GZEW: German ZEW Economic Sentiment
-
GIFO: German IFO Business Climate Index
-
GUC: German Unemployment Rate / Employment Change