CHAPTER 5

DISCUSSIONS AND CONCLUSION

5.1. Discussions
We have introduced some aspects of Bayesian decision theory that should
serve as a useful tool for engineers and decision makers. These methods seem to have
found favor in the economic and Business communities, partly because loss functions
seem more straightforward to them. In Bayesian methods, whereas there is the
problem of prior probabilities (i.e., unknown or vaguely defined, and the use of
subjectivity), these probabilities can revise on receipt of additional data and
consequent decision making is being implemented in one way or another in spite of
the many uncertainties faced. The decision maker learns from the analysis and design,
makes additional experiments and gathers more information. According to this
procedure, the decision maker can do less and less subject to error.
In this thesis, Bayes and minimax rules have been considered. We can also see
that the weakness of the minimax rule is intuitively apparent. It is very conservative
procedure that places all its emphasis on guarding against the worst possible case. In
fact, this worst case might not be very likely to occur. To make this idea more precise,
we can assign a probability distribution to the state of nature. This distribution is
called the prior distribution which is used in Bayesian method. We extend Bayes’ rule
to situations in which a posterior distribution is estimated on the basis of observations
or on the basis of known data (sometimes assumed). In such cases, it can be justifiable
to apply subjective probabilities quantifying personal knowledge or belief.
When we compute the minimax rule, we get d2 is the minimax solution. Then,
we compute the Bayes’s rule on the basis of previous events and from actual-
lengthening data by using the risk functions which have been calculated from the
minimax rules and the prior distribution from the forecasting of expert engineer’s
degree of belief. From Equation 4.3, we can see that d2 is the Bayes’s rule by
comparing the Bayes risk numbers. This Bayes’s rule is less conservative than the
minimax rule in which the action a1 (1230 cm height) is chosen from observation x2
 74

(1255 cm height) since the prior distribution for this Bayes’s rule puts more weights
on . According to this way, the action (1230 cm height) is chosen from
observation x1 (1230 cm height) and action a2 (1280 cm height) is chosen from
observation x3 (1280 cm height). If the prior distribution were changed sufficiently,
the Bayes’s rule would change. After have been calculated the posterior risk with
observed data X = x2 = 1255cm height (the additional information or sampling
information); we choose the smaller risk function among these two (PR) functions
(for two actions). It is the required action for the optimal decision d2.
To do this, Bayesian decision method leads to an optimal decision considering
the expected loss of all possible values of the random parameters by posterior
distribution. The object in decision making is to choose an action that minimizes the
expected values of the loss function with respect to the posterior distribution, if data
are available. However, if data are not available, the expected loss should be
minimized with respect to the prior distribution. Whenever we have the updated or
new information and we want to check the probability error of observed data in their
relative resource, Bayesian decision rule and the minimum error rate classification can
be applied to have a terminal decision. These rule and classification are able to get the
least value of probability error to be awareness of the good accuracy.
We have seen that the decision reliability of the efficiency in the flood
occurrence of Ayeyarwady River in Sagaing region for a newly constructed the
protective embankments cases has been determined with an error of 20 percent which
is calculated by using Equation 3.20 in terms of posterior error and prior predicative
density functions. Similarly, the probability error is calculated by using Equation 3.22
in terms of class conditional likelihood functions and Jeffrey’s prior (0.5,0.5). Based
on these situations, the reliability of our project can be revised whether it is
appropriate procedure or not in the future by Bayesian method. Thus, the reliability of
the protective embankments cases is predicted to be 95% if it is found the failure
twice. In an average reliability of 95%, this method can be computed the number of
continuously successful tested samples n = 18.

5.2. Conclusion
In this study, we see that, the concept of probability can be defined as the
degree to which a person believes a proposition and Bayes’ theorem can be used as a
rule to infer or update the degree of belief in light of new information.
 75

As a result, the posterior Bayes’ rule is to build an embankment of height 1230

cm because of its smaller posterior risk. This reflects a reduction in the uncertainty.
According to this way, we get the Bayes optimal decision and when this is adopted,
the decision maker expects to make a profit of $2,450 per embankment on average.
Similarly, in Bayesian Decision rule and minimum error rate classification
approach, the optimal total risk or Bayes’ risk gives $2450. This strategy is the
optimal Bayes’ decision strategy. The result of this optimal decision strategy is the
same as decision when it is carried out by minimax rules or Bayes’ rule.
Moreover, the conditional risk function and minimum risk decision rule are classified
to be more precise. Thus, decision maker can choose decision for the protection
of flood events even if the maximum water level of Ayeyarwady River will be
reached 1274 cm in the future.
After have been considered the Bayesian revision of reliability assessment, our
decision system of the construction of embankments case is not failed in the future
because the expected value of beta distributed reliability estimate is averagely 90%
and the required number of consecutively successful test maximum water levels are
eight samples. The corresponding standard deviation is 0.008. Thus, we have done the
terminal decision for the protection of the flood event. And also Bayesian method can
be revised our problem after it is completed. It is a useful method to update criteria
used after receiving the additional information. Thus, we can conclude that Bayesian
decision method gives the optimal decision strategy. This method has the physically
powerful in theoretical backdrop and the reasonably useful for a decision maker.
This thesis is presented on practical experiences in the gauging Sagaing station
(with reference data, which is an actual-lengthening water levels and flood events) to
be insights into the flood occurrence of Ayeyarwady River and protective
embankments cases. After have been studying about these events by Bayesian
decision method, these trends are concerned with the environmental impacts. In fact,
Bayesian decision method is served as a useful tool for Civil and Environmental
engineers in the real world.