Public defence in Systems and Operations Research, M.Sc. (Tech) Pekka Laitila

Risk management and decision-making under uncertainty are common challenges in business and public administration. Often the framework of a decision-making problem consists of various types of factors and variables whose mutual probabilistic dependencies may be difficult to know or perceive exactly. For instance, there might not be suitable historical data available, or the relevant data may be difficult to identify. These problems are typical in situations where risks are novel or unprecedented. Among such instances are, e.g., unique projects, ecological and economical disasters, and governmental conflicts.

Even though there might be a lack of suitable historical data, there is often an abundance of expert insight available, along with diverse information on indirectly related factors. In these situations, analysis of risks and decision-making under uncertainty can effectively be supported by Bayesian networks (BNs). A BN represents a system of linked components both visually and numerically enabling a rigorous quantification of risks and a clear communication of the components’ interaction. BNs can be constructed based on various information sources such as experimental data, historical data, and expert knowledge. The applications of BNs are numerous and cover a wide range of domains, such as medical decision support, risk analysis concerning epidemics, ecosystems, and industry, as well as policy and military planning.

The dissertation elaborates the construction of BNs by expert elicitation which involves subjective assessments of a domain expert and is often required in practical applications. The main contribution is the development of new elicitation approaches that help the expert to establish required numerical dependencies between BN components. The approaches improve an existing elicitation method commonly used in BN applications. They reduce the elicitation effort of the expert and also extend the application scope of the underlying method. Their practical execution is supported by thorough guidelines and online implementations. Consequently, the new approaches facilitate and promote the effective and diverse utilization of BNs in various applications.

Opponent is Professor Norman Fenton, Queen Mary University of London, UK

Custos is Professor Kai Virtanen, Aalto University School of Science, Department of Mathematics and Systems Analysis

Contact details of the doctoral student: [email protected]

The public defence will be organised on campus and via Zoom. Link to the event

The doctoral thesis is publicly displayed 10 days before the defence in the publication archive Aaltodoc of Aalto University.

Electronic thesis