B. Cross validation
After selecting the structure of model, the experimental
design is implemented. The experimental data are divided
into two groups call estimating data and validation data. To
improve the accuracy, cross validation are performed by
using different information and the validation results are
averaged over the rounds. In this paper various type of cross
595
validation are studied. Then advantage of each method has
been applied to system. In this research, 6 times-sampling
data sets cover the condition of system is experimented for ensure the accuracy of validation. Then data sets are divided
for training and validation. The training data are established by cascade all subsample data into the one group. The
validation data is obtained by expands another one set from the training data. Then system identification is operated.
Finally, a suitable model in any condition from cross validation technique is obtained.