EURO Journal on Decision Processes最新文献

The volume and availability of business and finance data may continue to increase in the near future. However, the utility of such data is by no means straightforward due to a lack of integration between data-driven techniques and usual decision-making processes. This paper aims to integrate data with multiobjective decision-making in cash management by means of machine learning. To this end, we first consider cash flow forecasting as a data-driven procedure to be used as a key input to multiobjective cash management problem in which both cost and risk are goals to minimize. Next, we compute the forecasting premium, namely, how much value can be achieved in exchange of predictive accuracy. Finally, we provide cash managers with a general methodology to improve decision-making in cash management through the use of data and machine learning techniques. This methodology is based on a novel closed-loop procedure in which the estimated forecasting premium (if any) is used as a critical feedback information to find better forecasting models and, ultimately, better cost-risk results in cash management.

Conditional preference networks (CP-nets) provide a compact and intuitive graphical tool to represent the preferences of a user. However, learning such a structure is known to be a difficult problem due to its combinatorial nature. We propose, in this paper, a new, efficient, and robust query-based learning algorithm for acyclic CP-nets. In particular, our algorithm takes into account the contradictions between multiple users’ preferences by searching in a principled way the variables that affect the preferences. We provide complexity results of the algorithm, and demonstrate its efficiency through an empirical evaluation on synthetic and on real databases.

Since the early applications of operations research and management science techniques in corporate financial management, financial decision support has evolved to a multi-disciplinary field combing different analytical approaches and technologies for supporting the decision-making process for financial problems faced by firms, organizations, and individuals. This paper provides an overview of the nature of financial decision support and its contributions, covering past developments and advances, as well as current trends and emerging topics on methodological, application, and implementation issues.

Biases like overconfidence and anchoring affect values elicited from people in predictable ways—due to people’s inherent cognitive processes. The more-or-less elicitation (MOLE) process takes insights from how biases affect people’s decisions to design an elicitation process to mitigate or eliminate bias. MOLE relies on four, key insights: (1) uncertainty regarding the location of estimates means people can be unwilling to exclude values they would not specifically include; (2) repeated estimates can be averaged to produce a better, final estimate; (3) people are better at relative than absolute judgements; and, (4) consideration of multiple values prevents anchoring on a particular number. MOLE achieves these by having people repeatedly choose between options presented to them by the computerized tool rather than making estimates directly, and constructing a range logically consistent with (i.e., not ruled out by) the person’s choices in the background. Herein, MOLE is compared, across four experiments, with eight elicitation processes—all requiring direct estimation of values—and is shown to greatly reduce overconfidence in estimated ranges and to generate best guesses that are more accurate than directly estimated equivalents. This is demonstrated across three domains—in perceptual and epistemic uncertainty and in a forecasting task.

This paper develops an R&D decision-making model in the real options framework. The model is generic enough to capture three types of uncertainty in an R&D project, namely, uncertainty of research duration and costs, market value of technology, and a competitor’s technology development. I derive analytical solutions, which help practitioners and researchers to evaluate various cases of R&D investment. Further, by analyzing the model with a wide range of parameter values, I reveal the following effects of the three types of uncertainty on R&D investment: higher uncertainty of research duration and costs, unlike market value uncertainty, speeds up investment, especially combined with a higher risk of competition. The investment timing can be U-shaped in the strength of competition because of the trade-off between the preemptive investment effect and the decreased project value effect. These results can account for empirical findings about the uncertainty–investment relation in industries with high R&D intensity and severe competition.

The occurrence of unexpected events that are extreme in magnitude, rare in frequency, and short-lived in duration poses distinctive challenges to decision makers and planners. In this paper we examine the impact of negative versions of these events, which we term “shocks”, on the judgmental forecasts of subjects experiencing them. A behavioral experiment asking participants to forecast monthly time series in the presence of temporary but extreme decreases in those series is used. Average changes to annual prediction intervals and 1-month ahead forecasts were much smaller than the magnitude of the shock and occurred in proportion to the size of the shock. Changes to prediction intervals were more persistent for moderate than large shocks, and larger for shocks occurring a second time. Our results provide supporting evidence for the view that decision makers underweight rare and extreme events rather than overweight them, consistent with a discounting or forgetting effect. The behavioral findings are relevant to operations researchers involved in expert judgment elicitation and in supporting decision making.

In a search for a positive model of decision-making with observable primitives, we rely on the burgeoning literature in cognitive neuroscience to construct a three-element machine (agent). Its control unit initiates either impulsive or cognitive elements to solve a problem in a stationary Markov environment, the element chosen depends on whether the problem is mundane or novel, memory of past successes, and the strength of inhibition. Our predictions are based on a stationary asymptotic distribution of the memory, which, depending on the parameters, can generate different “characters”, e.g., an uptight dimwit, who could succeed more often with less inhibition, as well as a laid-back wise-guy, who could gain more with a stronger inhibition of impulsive (intuitive) responses. As one would expect, stronger inhibition and lower cognitive costs increase the frequency of decisions made by the cognitive element. More surprisingly, increasing the “carrot” and reducing the “stick” (being in a more supportive environment) enhance contemplative decisions (made by the cognitive unit) for an alert agent, i.e., the one who identifies novel problems frequently enough.

Quantitative risk assessments for offshore oil and gas installations have been developed and used to support decision making about major hazards risk for more than 30 years. Initially, these studies were used to support the design process, aiming to develop installations that could be operated safely throughout their lifetime. As installations were put into operation, the studies were updated with as-built and operational information to provide a basis for making decisions also in the operational phase. This was however only partially successful, and the general impression has been that the studies have not been very actively used in operations. Many explanations have been given, the most common being that the reports were too complicated and written for risk analysis experts, not operations personnel on offshore installations and that the results could not be updated sufficiently often to reflect changes in risk on a day-by-day basis. This may be a part of the explanation, but in this paper, we have looked into the decision context and the types of decisions made in operation, compared to those in the design phase. Based on this, it is concluded that the focus of existing models need to be extended to cover activity risk in a more detailed way, as well as the risk associated with the technical systems. Instead, a revised methodology for developing quantitative risk assessments is proposed, focusing on the parameters and activities that change during operation. The methodology has also been tested on an offshore installation, to investigate the feasibility in practice.