Geology Reference
In-Depth Information
Fig. 6.14 Scatter plots of conjugate gradient ANN. a training, b validation
value is worse (smaller) during validation than during calibration for all of the
models, as is expected and was much smaller for conjugate type ANN. The AR-
MAX models outperform all ANN models as measured by these two statistics with
values of 0.89 and 0.90 respectively during training. Even though most of the
statistic values were better for ARMAX model, the MBE was found better in case
of ANN model with Levenberg
Marquardt algorithm. The scatter plots of different
ANN models in this study during training and validation are given in Figs.
6.14
,
6.15
,
6.16
. The prediction results of ARMAX model during the training period are
given in the Fig.
6.17
.
-
Fig. 6.15 Scatter plots of ANN-BFGS. a training, b validation
