Time Series Forecasting: Japanese Yen

For this project, I used time series forecasting and linear regression modeling to predict future movements in the value of the Japanese yen versus the U.S. dollar.

Time-Series Forecasting

In the time_series_analysis notebook, I loaded historical Dollar-Yen exchange rate futures data and applied time series analysis and modeling to determine whether there is any predictable behavior.

Based on the plot of "settle" prices of yen futures, we can see a long-term strengthening of the Japanese Yen against the Dollar. There do seem to be some more medium, 1-3 year consistent trends, but on a daily basis, there are a lot of short-term ups and downs.

Decomposition using a Hodrick-Prescott Filter

Smoothing with the HP Filter and plotting the resulting trend against the actual futures returns, we can see that there's a lot of short term fluctuations that deviate around this trend. Perhaps these would represent profitable trading opportunities: For example, when the blue line deviates far below the orange, we can see this as a sign that the Yen is temporarily more undervalued than it should be (and, therefore, we'd see this as a short-term buying opportunity).

Forecasting Returns using an ARMA Model

I used an ARMA model to forecast returns and determined that the model is not a good fit based on all the p-values being greater than .05.

Forecasting the Settle Price using an ARIMA Model

I used an ARIMA model to forecast returns and also determined that the model is not a good fit based on all the p-values being greater than .05. The ARIMA model forecast shows that the Japanese Yen will increase at a linear rate in the next five days.

Forecasting Volatility with GARCH

The Garch model seemed to be good fit for predicting volatility because most of the p-values are below .05.

Conclusions

• Based on my time series analysis, I would buy the yen when the settle price is below the trend line.
• The risk of the yen is expected to increase within the next five days.
• Based on the model evaluation, I do not feel confident in using these models for trading due to their p-values being higher than .05.

Linear Regression Forecasting

In the regression_analysis notebook, I built a Scikit-Learn linear regression model to predict Yen futures ("settle") returns with lagged Yen futures returns and categorical calendar seasonal effects (e.g., day-of-week or week-of-year seasonal effects).

The model performs better with in-sample data (training data) because it has a RMSE of .415 as opposed to the RMSE of .565 of the out of sample data.