Multi-Actor Portfolio Analysis Methodology: A Stochastic and Heuristic based approach to deal with information scarce environments

Abstract

Previous academic work on Portfolio Decision Analysis (PDA) have pointed at the advantage of using PDA in multi-actor situations and environmental sectors, but have also pointed at the lack of research done in such case studies. Currently there exist situations where PDA would be beneficial but current methods are not able to analyse the systems. This is because most of the PDA and MCDA methods require certain data that is often absent in real life case studies, such as reliable information on alternatives performances including synergies and the availability of time and willingness for detailed preference elicitation. Multi-actor situations make preference elicitation an even more time consuming aspect. Therefore, this thesis develops a novel methodology making PDA analysis accessible, robust and fast in decision support to multi-actor systems lacking detailed synergy and preference information. First the portfolios are created in a linear-additivity framework covering the solution space (creating thus all feasible portfolios). The methodology uses different strains of thought and concepts within the MCDA community, such as a combination of stochastic and heuristic approaches to analyse the portfolios. The stochastic techniques include Monte Carlo sampling for uncertainty propagation of the alternative performances and stochastic multicriteria acceptability analysis (SMAA) to handle attribute weight uncertainties. The heuristics applied stand within the Fast and Frugal Heuristics tradition (FFH). The studied heuristics are the Take-The-Best Heuristic and a Lexicographic heuristic to incorporate synergy information. Application results in portfolio shortlists for the involved actors. In order to analyze consensus among the actors a Core Indices (CI) approach is applied. Subsequently a Sobol sensitivity analysis is conducted on the most promising portfolios. The integration of these methods result in an overarching unconventional approach capable of informing decision makers on most promising portfolios and most important attributes. An application of this methodology is shown through application to both simulated data and a real life case study in a multi-actor case-study in India, on wastewater treatment, reuse and resource recovery. The case study encompasses multiple stakeholders in which, due to the political nature of the situation, it is difficult to obtain explicit preference. The alternatives also likely have interactions and interdependencies, affecting the portfolio set performance. The application shows that the method is able to deal with the complexities of real life case studies. For both the simulated data set as for the case study consensus on portfolios, alternatives and corresponding attributes have identified for further decision-making. Regarding the analysis of the attribute weights uncertainty the SMAA-O approach is deemed better than the Take-The-Best approach if a complete ordinal ranking of the attributes is available. Furthermore, the approach of lexicographic heuristics to deal with the lack of detailed synergy information seems promising for situations where synergy information is lacking. For the general application of the method there are, however, some limitations due to assumed linear properties, these encompass: The assumed linear additivity of the alternatives in the portfolio creation, the monotonic behaviour of the marginal value functions and the hierarchical addition aggregation value function. These form also the points where the method can be expanded upon to integrate these assumptions and corresponding uncertainties. In the future the method can be expanded upon in the open source software to make further use of the computing power able to perform stochastic analysis and to build upon the novel application of heuristics, quickly allowing PDA analysis in ever more environmental case studies.