Gérard Colson

The OR's prize winner and the software ARGOS: how a multijudge and multicriteria ranking GDSS helps a jury to attribute a scientific award

Abstract An annual prize is attributed to the best students final work by the Belgian Operations Research Society. In 1992, the author happened to be a member of the jury designated by this Society to evaluate the candidates and to elect the best one. The three others members of the jury were kindly convinced by the author to use a multiple criteria and multijudge procedure to discriminate the eight candidates qualities, parallely to their more classical global evaluation. In order to facilitate the procedure, a group decision support system (GDSS) was fed by four evaluation matrices given by the team, who defined a common set of criteria completed by an overall direct evaluation. The purposes of this paper are to describe this real-world experiment and to present a multiple criteria procedure of candidates’ selection and ranking supported by the software ARGOS (acronym of aid to the ranking to be made by a group of decision makers using an outranking support ). This GDSS allows a team of decision makers to run one or several releases of ELECTRE and PROMETHEE methods of preference aggregation in order to deal with the problems of sorting, ranking and of electing the best candidate. A first release of ARGOS is able to deal with the two phases of a group decision process: the multiple criteria preference aggregation performed by each member of the group in the first phase, and, in the second phase, the comparison of results issued from the application of several functions of social choices for supporting the deliberation process of the group, e.g. Borda or Condorcet functions, the Raynauds prudent orders and some rank disagreements functions. In its second release, ARGOS has been enriched by the addition of JUDGES, the name for another GDSS which compares the rankings given by the teams members and the candidates’ distributions of ranks, provides advices for seeking a consensus, and supports voting simulations. Scope and purpose Multiple criteria literature has a rich history. Outranking approach is a very dynamical branch of this literature with its ‘star methods’ Electre and Promethee. More and more often, real-world applications of these methods involve small teams of decision makers. Another branch of decision aid is concerned with facilitation of small group meeting in view of ranking projects or candidates, and several softwares have been designed for this purpose. Few of them are focused on the junction of both approaches allowing to deal with outranking in the framework of small team decision making. The purpose of this paper is to present the software ARGOS, which addresses this latter problem through its use in the facilitation of the attribution of a scientific award.

An integrated multiobjective portfolio management system

The literature on portfolio management is rather abundant, with many valuable suggestions for improving the activity of large or small companies involved in the investment business. Few propositions, if any, are concerned with a full integrated system for helping portfolio managers in the considerable tasks of collecting information, making predictions about the market and about specific securities behaviours, on the one hand and, on the other hand, of compromising multiple conflictual objectives pursued by the investment committee, and controlling the complete activity of the service in charge of the sequence of tasks. The authors present the first global computerized system for managing, in an integrated way, the three phases involved in all portfolio management: the information phase; the decision phase; and the control phase (including the evaluation of performances and effectiveness of the system). The core of the system is the junction of the two main processing models. The single decision model (SDM) makes use of the Bayesian approach to process the information and to combine the forecasts supplied by internal analysts to the messages issued from external correspondents. The outputs of the SDM are eight criteria for evaluating each security scrutinized by this company. The simultaneous management model (SMM) is fed by these outputs and by the decisional parameters; it revises the current portfolio on the basis of a multiple criteria evaluation. Efficient compromises are obtained here by goal programming. In this paper, the control phase is not detailed, while an experiment of the system in a small company is reported.

Editorial: Models and methods in multiple objectives decision making

Publisher Summary This chapter discusses models and methods in multiple objectives decision making. Multicriteria decision making (MCDM) is a world of concepts, approaches, models, and methods to help the decision makers to describe, evaluate, sort, rank, select, or objects, candidates, products, projects, etc. on the basis of an evaluation expressed by scores, values, and preference intensities according to several criteria. These criteria may represent different aspects of the teleology, namely, objectives, goals, targets, reference values, and aspiration levels, utility. An attribute is a characteristic of an object that can be evaluated objectively or subjectively by one or several persons according to a measurement scale; different measurement scales and modalities can be used. According to Colson, the uncertainty arises from a lack of knowledge of some aspects of the decision problem. The main factors of uncertainty are the role of chance, the ambiguity, and an insufficient understanding of the reduction of complexity to accept to solve a practical problem.

A public warehouses selection support system

Abstract Our software public warehouses selection support (PWSS) has two purposes: to select public warehouses according to several criteria and to exploit a database when some data are missing. It has been applied to a database of 280 Belgian enterprises, which are working in the fields of road transportation and logistics services providers. The database is organized as a four entries table: on rows, the enterprises are clustered by provinces and regions, and, on columns, items of information about the enterprises’ warehouses are clustered into general topics. Such items are treated like choice criteria and topics like criterion families: i.e. buildings, customs, logistics and handling. Thanks to software PWSS, the user can make a multiple criteria selection of a subset of public warehouses fitting as well as possible his needs and preferences. He 1 can obtain too a multiple criteria ranking of warehouses based on a satisfaction index and a confidence index. This latter index measures his confidence in the results of his satisfaction index, its value depending on the importance he assesses to criteria with non-available data in the explored subset.

Mars: A multiattribute utility ranking support for risk situations with a “P,Q,I,R” relational system of preferences

The computer programs for utility assessments present some drawbacks due to the difficult questionning by the decision-maker, and mainly to an exaggerated requirement in the precision of the answers. The main originality of MARS, a support for ranking and evaluating uncertain prospects by means of the subjective expected utility theory, is the addition of the relational system of preferences (P,Q,I,R) by Roy and Vincke into a classical interview process proposed by the multiattribute utility theory. This allows the acceptance of imprecise answers and the treatment of ambiguity in the interview process. Moreover, a degree of confidence in the independence conditions among attributes, and general propensities for aversion or preference towards risk are computed. MARS contains two submembers: PUMA and UPEP. The set of programs PUMA (program for utility with multiple attributes) firstly determines the degree of confidence in a mutual utility independence or in an additive independence between the attributes, and advises a specific form of the utility function; other subsets of PUMA help construct the uniattribute utility functions, scale the constants and, finally, rank the uncertain prospects in a decreasing order of expected multiattribute utility. The set of programs UPEP (utility and probability encoding program) sensitizes the user to his risk attitude and computes his personal utility function in a confidence band. It also allows the construction of a univariate continuous distribution of subjective probabilities and its adjustment by a beta distribution.

Toward a bipolar theory of risk

Abstract Every mans attitude toward risk is a priori characterized by both a speculative pole and a conservative pole. Most men like high (speculative) gains and simultaneously dislike losses associated with a given venture. Such an attitude generates a conflict, which must be resolved when dealing with decisions under uncertainty. The traditional theories of risk implicitly assume that this conflict is erased when fitting an everywhere-concave (or convex) utility function. Such unipolar theories should be compared with a bipolar theory and more generally with a multicriterion theory of risk. Our mean-bipolar risk vector analysis generalizes the traditional mean-risk analysis. In Portfolio Selection, the mean-variance approach can be extended to include the speculative pole as well as the conservative pole: the variance is viewed as both desired and undesired components of riskiness.

Multiple Criteria and Multiple Periods Performance Analysis: The Comparison of North African Railways

Multi‐period differences of technical and financial performances are analysed by comparing five North African railways over the period (1990–2004). A first approach is based on the Malmquist DEA TFP index for measuring the total factors productivity change, decomposed into technical efficiency change and technological changes. A multiple criteria analysis is also performed using the PROMETHEE II method and the software ARGOS. These methods provide complementary detailed information, especially by discriminating the technological and management progresses by Malmquist and the two dimensions of performance by Promethee: that are the service to the community and the enterprises performances, often in conflict.

Decision Analysis in Entrepreneurship: how a New Venture Creation can Benefit from it

The entrepreneurship process encounters a crucial step in the decision of launching or not the new venture, when the business plan has been approved. This crucial decision could be better supported by a formal risk assessment, even if this step is often fed by a subjective estimation assessed by the entrepreneur. The Decision Analysis (DA) model is useful for the “one shot” strategic decision of launching or not a new business venture. This paper is a real-world application of the DA approach to small entrepreneurship, following the DA guidelines proposed by the Stanford Research Institute in the seventies for big enterprises or organisations. Through a real world case study, this paper uncovers the advantages and the inconveniences of the DA model applied to small new ventures.