Seismic risk analysis using Bayesian belief networks - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 183

Liquefaction

s¢ vo

q c a max D 50

s vo

M

7.6 Bayesian network for liquefaction assessment.

Regional damage
estimation

Evaluation of a
group of buildings

Evaluation of
individual
buildings

7.7 Increase in level of details for regional and individual building
damage estimation.

with fi xed structure achieved the best classifi cation results in terms of the
criteria considered. The compatible algorithm is hill climbing, which achieves
similar correct and incorrect classifi cation rates as the model with fi xed
structure. However, performance of the other criteria is lower.
Table 7.5 gives information on how the BBN models work for the lique-
faction status, i.e., 'yes' and 'no'. This table can be used to explore the
trade-offs among BBN models over different scenarios. For example, the
model with hill climbing algorithm has a larger TP rate than the model with
fi xed structure for liquefaction 'Yes', but it also has a larger FP rate at the
same time.

7.4 Regional damage estimation

Damage estimation of buildings requires different spatial resolutions. The
damage estimation can be undertaken for a city, group of buildings or indi-
vidual buildings. With each assessment type, the level of details and infor-
mation required increases as shown in Fig. 7.7.