MAIPARK Vulnerability Analyzer is a multi-peril risk analysis platform developed to support quantitative assessment of building fragility and vulnerability across different hazard types. The platform is designed as a global analytical framework, where individual hazard modules, such as earthquake and flood, are implemented through dedicated analysis menus.
The analytical models are calibrated using historical insurance loss and damage data obtained from PT Reasuransi MAIPARK Indonesia. These datasets provide empirical evidence of observed damage, enabling the tool to reflect real-world risk characteristics relevant to insurance and reinsurance applications.
The development of the MAIPARK Vulnerability Analyzer follows a structured methodology that integrates Machine Learning and statistical damage modeling. Historical exposure and damage records are first processed and divided into training and testing datasets to ensure objective performance evaluation.
In the initial modeling stage, a Random Forest Classifier is applied to learn complex, non-linear relationships between hazard intensity, building characteristics (such as construction age, number of stories, and structural attributes), and discrete damage states ranging from None to Complete. This stage serves to evaluate classification accuracy, robustness, and feature relevance.
Following the Machine Learning training–testing phase, the damage classification results are used as the basis for probabilistic damage modeling. An Ordinal Logistic Regression framework is employed to explicitly account for the ordered nature of damage states, ensuring a consistent representation of progressive damage severity as hazard intensity increases.
Using the ordinal regression outputs, the tool constructs Fragility Curves, which describe the probability of exceeding specific damage states, and Vulnerability (Loss Ratio) Curves, which translate damage probabilities into expected loss ratios. For earthquake analysis, these curves are derived using Modified Mercalli Intensity (MMI) and Peak Ground Acceleration (PGA) as seismic intensity measures.
This modular modeling concept allows the platform to be consistently extended to other perils, such as flood, while maintaining a unified methodological foundation for risk quantification and decision support.
The development of the MAIPARK Vulnerability Analyzer is the result of collaborative efforts with the following roles:
For technical inquiries, collaboration, or further information, please contact:
Contact Person:
Sarah Eden Suryanto
📧 sarah.eden@maipark.com