Civil Engineering Reference
In-Depth Information
Fig. 3.3
Real PMV index and neural network approximation for the summer data VA1a (
top
)and
VA 1 b (
middle
), and absolute error (real PMV—neural network approximation) for both data sets
(
bottom
)
neural network model when computing the PMV index since in the worst case they
have a relative error equal to 4.7%.
On the other hand, with respect to winter data sets, the neural network model
obtains a RMS error value of
e
RMS
VA
2
a
0145 and
e
RMS
VA
2
b
0123, for validation
data sets VA2a, VA2b, respectively, see Fig.
3.4
. As for the summer case, these results
are inside the expected range with a relative error in the worst case equal to 5.9%.
Moreover, they prove the efficiency of the neural network model to calculate the
PMV index with real data from both non-working and working days.
=
0
.
=
0
.