b'Support Vector AnalysisThis research improves risk quantification, reduces human error in risk for Computational Riskassessment, and enables subject matter experts to focus on critical risk areas.Assessment, Decision-Making,M odern probabilistic risk assessment is a method of documenting the discoveries of an analyst based on their qualifications and available and Vulnerability Discovery insystem information. This is not a process to suggest new, previously Complex Systems unknown risk. The purpose of this research was to develop methods of auto-detecting possible vulnerabilities in system designs to uncover previously unseen issues and reduce human error/costs by allowing analysts to focus on critical areas via intelligent, efficient sampling of the systems parameter space.The objective was to develop a fundamentally new way of exploring supervised binary classification using support vector machines to intelligently guide the TOTAL APPROVED AMOUNT:sampling process through very high-dimensional parameter spaces to analyze $125,000 over 1 year logical flaws in the systems complex design. Currently, the machine learning field PROJECT NUMBER:is dominated by pattern recognition, data representation, and forecasting. Research 20A1054-008 in machine learning techniques to discover logical fallacies is lacking. The newly developed methodology uses machine learning to develop a computational method PRINCIPAL INVESTIGATOR:for analyzing logical constructs (e.g., fault trees) represented as Boolean expressions.Andrei GribokThis project produced two primary outcomes. This first is a new, broadly CO-INVESTIGATOR: applicable methodology that uses intelligently guided space sampling methods Curtis Smith, INL to drastically reduce the number of system configurations needing to be analyzed. This methodology allows researchers to auto-detect possible vulnerabilities in system designs, devices, and networks to uncover previously unseen issues and reduce human error and costs by enabling analysts to focus on critical areas via intelligent, efficient sampling of the systems parameter space. The second outcome is a demonstration of the benefits of machine learning, as applied to a case study relating a computational risk assessment applied to probabilistic risk assessment.Risk assessment gives insights into potential system vulnerabilities. Historically modeled by human experts and represented via fault trees (logic model for the system). However, the process and models are complex. The number of possible paths to failure for a system with 100 components is 1,267,650,600,228,229,401,496,703,205,376. 129'