Jeffrey L. Ringuest

A multiobjective evolutionary approach for linearly constrained project selection under uncertainty

In the project selection problem a decision maker is required to allocate limited resources among an available set of competing projects. These projects could arise, although not exclusively, in an R&D, information technology or capital budgeting context. We propose an evolutionary method for project selection problems with partially funded projects, multiple (stochastic) objectives, project interdependencies (in the objectives), and a linear structure for resource constraints. The method is based on posterior articulation of preferences and is able to approximate the efficient frontier composed of stochastically nondominated solutions. We compared the method with the stochastic parameter space investigation method (PSI) and illustrate it by means of an R&D portfolio problem under uncertainty based on Monte Carlo simulation.

Multiobjective optimization : behavioral and computational considerations

1. Introduction.- 1.1 Multiple-Objective Optimization.- 1.2 Dominance And Efficiency.- 1.3 Multiattribute Value And Utility Theory.- 1.4 Functional Forms And Independence Conditions.- 1.5 Value Functions As Compared To Utility Functions.- 1.6 Optimizing The Multiattribute Utility Or Value Function.- 1.7 References.- 1.8 Other Relevant Readings.- 2. Linear Goal Programming.- 2.1 The Goal Programming Model.- 2.2 Aspiration Levels.- 2.3 Weights.- 2.4 Preemptive Priorities.- 2.5 Multiattribute Value Theory.- 2.6 Specifying The Weights In An Additive Value Function.- 2.7 Sensitivity Analysis.- 2.8 References.- 2.9 Other Relevant Readings.- 3. Generalizing Goal Programming.- 3.1 Linear Goal Programming.- 3.2 Piecewise Linear Approximations Of Single Attribute Value Functions.- 3.3 Goal Programming With A Multiplicative Value Function.- 3.4 Nonlinear Goal Programming.- 3.5 References.- 4. Compromise Programming.- 4.1 Ideal Solutions.- 4.2 Compromise Functions.- 4.3 Compromise Solutions And The Compromise Set.- 4.4 The Anti-Ideal And Compromise Programming.- 4.5 The Method Of The Displaced Ideal.- 4.6 Compromise Programming, Linear Goal Programming, And Multiattribute Value Functions.- 4.7 References.- 5. Decision Making and the Efficient Set.- 5.1 The Efficient Set.- 5.2 Intra-Set Point Generation.- 5.3 Filtering.- 5.4 Clustering.- 5.5 Matching And Grouping.- 5.6 Sectioning.- 5.7 A Stochastic Screening Approach.- 5.8 References.- 5.9 Other Relevant Readings.- 6. Interactive Methods.- 6.1 The General Interactive Approach.- 6.2 Examples Of Interactive Methods.- 6.3 Simplified Interactive Multiple Objective Linear Programming (SIMOLP).- 6.4 Interactive Multiobjective Complex Search.- 6.5 Choosing An Interactive Method.- 6.6 References.- 7. Computational Efficiency and Problems with Special Structure.- 7.1 Network Flow Problems.- 7.2 Multiple Objective Network Flow ProbLems.- 7.3 A Network Specialization Of A Multiobjective Simplex Algorithm.- 7.4 Compromise Solutions For The Multiobjective Network Flow Problem.- 7.5 Interactive Methods For The Multiobjective Network Flow Problem.- 7.6 References.- 8. Computational Efficiency and Linear Problems of General Structure.- 8.1 Computational Efficiency And The Ideal Solution.- 8.2 Test Problems.- 8.3 Computer Codes.- 8.4 Results.- 8.5 Other Computational Studies.- 8.6 References.- 9. Using Multiobjective Linear Programming to Reconcile Preferences Over Time.- 9.1 Preferences Over Time.- 9.2 The Behavioral Properties Of NPV.- 9.3 A More General NPV Model.- 9.4 Using Multiobjective Linear Programming As An Alternative To NPV.- 9.5 The Advantages Of Using Multiobjective Linear Programming For Reconciling Preferences Over Time.- 9.6 References.- 10. Data Presentation and Multiobjective Optimization.- 10.1 Data Representation And The Axioms Of Utility Theory.- 10.2 The Framing Of Decisions.- 10.3 Reconciling The Decision Frame.- 10.4 Perception Of The Ideal.- 10.5 References.

Interactive solutions for the linear multiobjective transportation problem

Abstract Mathematical programming problems that exhibit the mathematical structure of a transportation problem often arise in settings with multiple conflicting objectives. Existing procedures for analyzing these problems fall into two general categories. These methods either generate all nondominated solutions or they construct a single compromise solution. This paper presents two interactive algorithms which take advantage of the special form of the multiple objective transportation problem. Two examples are included to illustrate these algorithms and to demonstrate their viability.

Mean–Gini analysis in R&D portfolio selection

Abstract To date no single model has been published which fully satisfies the needs for a practical R&D project selection technique. Some earlier models cannot handle risk well, while others do not provide efficient portfolios. This paper will present a model, adapted from the literature of financial portfolio optimization, which provides a practical means of developing preferred portfolios of risky R&D projects. The method is simple and highly intuitive, requiring estimation of only two parameters, the expected return and the Gini coefficient. The Gini coefficient essentially replaces the variance in the two-parameter mean–variance model and results in a superior screening ability. The model that we present requires estimates of only these two parameters and, in turn, allows for relatively simple determination of stochastic dominance (SD) among candidate R&D portfolios. We apply our model to a simple artificial five-project set and then to a set of 30 actual candidate projects from an anonymous operating company. We demonstrate that we can determine the stochastically non-dominated portfolios for this real-world set of projects. Our technique, appropriate for all risk-averse decision makers, permits R&D managers to screen large numbers of candidate portfolios to discover those which they would prefer under the criteria of SD.

Formulating Optimal R&d Portfolios

OVERVIEW: R&D managers continue to report their dissatisfaction with currently-available R&D portfolio management models and the need for better and simpler modeling approaches. This article presents a simple, yet theoretically rigorous, method for designing optimal R&D portfolios, those that minimize risk for a given level of return. The model requires only the assumption that the relevant decision makers are risk-averse, a commonly accepted proposition. The model requires as input only an estimate of the probability of success for each project and an estimate of the financial return in case of success and failure. Finally, the model can be executed using only an ordinary EXCEL spreadsheet. The straightforward approach illustrated here should add considerable power and rigor to the design of optimal R&D portfolios.

Formulating R&D Portfolios That Account for Risk

OVERVIEW:R&D managers express displeasure with the state of the art in portfolio decision models and are often doubtful about their firms current portfolio of projects. Many feel that conventional portfolio decision models are impractical, requiring data that are almost impossible to estimate, or fail to take into account the risk-mitigating effects of diversification across a portfolio. The new approach presented here is scientifically rigorous, yet simple and practical. It can be implemented easily using only an EXCEL spreadsheet, and requires as inputs only the estimated success probabilities and payoffs of each R&D project. The model develops priorities for each of the R&D projects, which take into account the joint risk of the entire portfolio. The project priorities developed by this model are closely correlated with the internally established priorities for 54 R&D projects in two large ethical pharmaceutical firms.

Conditional stochastic dominance in R&D portfolio selection

This paper describes a methodology for the selection of research and development (RD however, they apply it using simulation in an R&D portfolio context. They apply the methodology to the portfolios of two actual companies and find that it generates priorities very close to those developed by internal company heuristics. They conclude that this methodology can be applied appropriately in these circumstances and that its recommendations are consistent with observed decision maker behavior. Their results suggest that an R&D manager should not consider project selection decisions in isolation, but, following this methodology, should take into account the context of the existing portfolio.

LP-metric sensitivity analysis for single and multi-attribute decision analysis

Abstract Analyzing the sensitivity of decisions to probability estimation error in single and multi-attribute problems and to errors in estimating additive multi-attribute value models in multi-attribute problems is an integral part of decision analysis. This paper presents an intuitive and tractable approach to this sensitivity analysis. Here a decision is considered insensitive if: 1) the probabilities or multi-attribute weights required for any other alternative to become preferred are not close to the original estimated probabilities and weights, and 2) the rank order of states implied by the probabilities or the rank order of attributes implied by the additive multi-attribute weights must change for any other alternative to become preferred. The sensitivity analysis is conducted using straight forward linear programming models. An example is used to demonstrate their application.

An application of multiattribute utility theory to the planning of emergency medical services

This research considers the problem of relating Emergency Medical Services (EMS) to patient outcome. The hypothesis is that response time alone may be misleading as an EMS performance criterion. This research uses methods for approximating multiattribute utility functions to consider both response time and on-the-scene care. The final result is an optimization problem where the response time and desired personnel requirements are decision variables. These are important inputs in the planning for Emergency Medical Services.

Interactive multiobjective complex search

Abstract An interactive procedure based on Boxs complex search is used to solve the vector maximization problem. This method has the advantage that the decision makers underlying value function need not be explicitly specified. Also, the problem may have nonlinear objective functions and nonlinear constraints. Several example problems are presented.