Bayesian network with quantitative input for maritime risk analysis

Abstract

This article presents an innovative approach towards integrating logistic regression and Bayesian networks (BNs) into maritime risk assessment. The approach has been developed and applied to a case study in the maritime industry, but has the potential for being adapted to other industries. Various applications of BNs as a modelling tool in maritime risk analysis have been widely seen in relevant literature. However, a common criticism of the Bayesian approach is that it requires too much information in the form of prior probabilities, and that such information is often difficult, if not impossible, to obtain in risk assessment. The traditional and common way to estimate prior probability of an accident is to use expert estimation (inputs) as a measure of uncertainty in risk analysis. In order to address the inherited problems associated with subjective probability (expert estimation), this study develops a binary logistic regression method of providing input for a BN, making use of different maritime accident data resources. Relevant risk assessment results have been achieved by measuring the safety levels of different types of vessels in different situations.

Keywords:

• Bayesian networks
• binary regression
• quantitative input
• risk assessment

Acknowledgement

This research was supported in part by research grants of J-BB7F & G-YJ48 from The Hong Kong Polytechnic University.

Notes

1. The value of secondhand ships was used as the loss (Table 6) once the total loss occurred.

2. Figure 3 illustrates identical changes to Figure 2 but with the addition of ‘Ship Condition’ set to 100% ‘Standard’ and 100% ‘Substandard’.

3. Figure 4 illustrates identical changes to Figure 2 but with the addition of ‘Classification Society’ set to 100% ‘IACS’ and 100% ‘Non-IACS’.

4. Figure 6 illustrates identical changes to Figure 2 but with the addition of ‘Ship Size’ set to 100% ‘Large’ and 100% ‘Small’.

5. Figure 7 illustrates identical changes to Figure 2 but with the addition of ‘Ship Age’ set to 100% ‘New’, 100% ‘Medium’ and 100% ‘Old’.