Fig. 2

Predictions of simulated data using various models and correlations by masking the last 14 data points. Our prediction with correlation uses 11 other correlated time series, and we made a prediction assuming no correlation with those time series. A training window of 120 time steps was used in each case. (a, c, e) use GBM and (b, d, f) use GOU. (a, b) use a correlation matrix with all off-diagonal elements of the correlation matrix equal to.4 with (c, d) using.6 and (e, f) using.8. Estimated parameters and correlations for these examples can be seen in Table 4 (Appendix 4)