João Paulo Costa

The AGAP system: A GDSS for project analysis and evaluation

Abstract This paper presents the ‘Aid to Groups of Analysis and evaluation of Projects’ (AGAP) system, a distributed group decision support system (GDSS) allowing multiple decision makers to cooperate in the evaluation and selection of investment projects. The system has a state of the art set of economic measures that can be set as criteria for use in several multicriteria decision aid methods. It supports both synchronous and asynchronous usage, and tries to enhance the communications and data sharing during asynchronous group meetings, providing support for decision at individual, inter-personal and collective levels. The system is described at a conceptual level, and compared to some other tools available to achieve the same aims.

Using ELECTRE TRI outranking method to sort MOMILP nondominated solutions

Abstract Several interactive methods exist to identify nondominated solutions in a Multiple Objective Mixed Integer Linear Program. But what if the Decision Maker is also interested in sorting those solutions (assigning them to pre-established ordinal categories)? We propose an interactive “branch-and-bound like” technique to progressively build the nondominated set, combined with ELECTRE TRI method (Pessimistic procedure) to sort identified nondominated solutions. A disaggregation approach is considered in order to avoid direct definition of all ELECTRE TRI preference parameters. Weight-importance coefficients are inferred and category reference profiles are determined based on assignment examples provided by the Decision Maker. A computation tool was developed with a twofold purpose: support the Decision Maker involved in a decision process and provide a test bed for research purposes.

A parallel implementation of the PROMETHEE method

The obtainment of a result from a decision support system is usually preceded by a structuring of the decision situation and followed by a robustness analysis phase. In this last phase the decision makers (DMs) observe the impact of changing the parameters of the decision model built during structuring, in order to learn about the situation and increase their confidence on the results. It is essential that the decision support system enables interactivity (i.e. provides short response times), otherwise during this phase the DMs will not be encouraged to be as exhaustive as required by the situation. This presents a computational challenge when the problems are not of trivial dimension. This paper discusses the application of parallel processing as a means to meet this challenge when building a decision support system based on the PROMETHEE multicriteria aid method. Several parallel programs have been built and compared on a 16-processor computer. Our purpose was to acquire some insight on how the parallel programs do perform under different situations and to identify the features of the method more relevant to its parallelisation. We verified that at some situations the reduction of the computers response time by means of parallel processing is quite appreciable and may foster the use of a decision support tool.

Computing non-dominated solutions in MOLFP

Abstract In this paper we present a technique to compute the maximum of a weighted sum of the objective functions in multiple objective linear fractional programming (MOLFP). The basic idea of the technique is to divide (by the approximate ‘middle’) the non-dominated region in two sub-regions and to analyze each of them in order to discard one if it can be proved that the maximum of the weighted sum is in the other. The process is repeated with the remaining region. The process will end when the remaining regions are so little that the differences among their non-dominated solutions are lower than a pre-defined error. Through the discarded regions it is possible to extract conditions that establish weight indifference regions. These conditions define the variation range of the weights that necessarily leads to the same non-dominated solution. An example, illustrating the concept, is presented. Some computational results indicating the performance of the technique are also presented.

An exact method for computing the nadir values in multiple objective linear programming

In this paper we propose a new method to determine the exact nadir (minimum) criterion values over the efficient set in multiple objective linear programming (MOLP). The basic idea of the method is to determine, for each criterion, the region of the weight space associated with the efficient solutions that have a value in that criterion below the minimum already known (by default, the minimum in the payoff table). If this region is empty, the nadir value has been found. Otherwise, a new efficient solution is computed using a weight vector picked from the delimited region and a new iteration is performed. The method is able to find the nadir values in MOLP problems with any number of objective functions, although the computational effort increases significantly with the number of objectives. Computational experiments are described and discussed, comparing two slightly different versions of the method.

An interactive method for multiple objective linear fractional programming problems

Abstract.Multiple objective linear fractional programming (MOLFP) is an important field of research. Using some branch and bound techniques, we have developed a new interactive method for MOLFP that drastically reduces the computational effort needed, while providing guidance for the decision maker in the choice of his/her preferred solutions. The basic idea of the computation phase of the algorithm is to optimize one of the fractional objective functions while constraining the others. Several linear programming problems, organized in a tree structure, are generated as the search evolves. The whole idea is simple and it results in a fast and very intuitive approach to exploring the non-dominated set of solutions in MOLFP, and eventually to finding the preferred solution.

Conflicting criteria, cooperating processors—some experiments on implementing a multicriteria decision support method on a parallel computer

Abstract This paper presents a step-by-step parallel approach to the ELECTRE III multiple criteria outranking method. Several implementations of the method were compared using a multiple instruction multiple data message-passing multiprocessor with 16 processing elements. The programs are described and results are presented considering several test situations. This allowed us to identify the implementation that ran faster and to study its behaviour. It was observed that under certain circumstances the speedups obtained by parallel versions were attractive, namely when applied to the rest problems in which the sequential program was slower.

A Multiple Reference Point Parallel Approach in MCDM

This paper presents a multiple reference point approach that enables oriented strategic search for non-dominated solutions of a multiobjective linear program.

Immune response elicited by an intranasally delivered HBsAg low-dose adsorbed to poly-ε-caprolactone based nanoparticles

Among new strategies to increase hepatitis B virus (HBV) vaccination, especially in developing countries, the development of self-administered vaccines is considered one of the most valuable. Nasal vaccination using polymeric nanoparticles (NPs) constitutes a valid approach to this issue. In detail, poly-ε-caprolactone (PCL)/chitosan NPs present advantages as a mucosal vaccine delivery system: the high resistance of PCL against degradation in biological fluids and the mucoadhesive and immunostimulatory properties of chitosan. In vitro studies revealed these NPs were retained in a mucus-secreting pulmonary epithelial cell line and were capable of entering into differentiated epithelial cells. The intranasal (IN) administration of 3 different doses of HBsAg (1.5 μg, 5 μg and 10 μg) adsorbed on a fixed amount of PCL/chitosan NPs (1614 μg) generated identical titers of serum anti-HBsAg IgG and anti-HBsAg sIgA in mice nasal secretions. Besides other factors, the NP surface characteristics, particularly, zeta potential differences among the administered formulations are believed to be implicated in the outcome of the immune response generated.

An algorithm based on particle swarm optimization for multiobjective bilevel linear problems

Improved MOPSO for multiobjective (at the upper level) bilevel linear problems.A hybrid scheme for the selection of the global best particles is proposed.An adaptive mutation is also introduced.The incorporation of these mechanisms led to significantly better results. This paper presents an improved multiple objective particle swarm optimization (MOPSO) algorithm to solve bilevel linear programming problems with multiple objective functions at the upper level. The algorithm aims to produce a good approximation of the entire Pareto front of the problem. We have previously designed a MOPSO algorithm for the same class of problems, in which several techniques for the global best selection were tested, including a new one. The algorithm revealed a good convergence towards the Pareto front but the diversity of the solutions was a drawback. The algorithm we propose herein uses a hybrid strategy for the global best selection and an adaptive mutation operator. The incorporation of these mechanisms led to an improved algorithm, which also showed better overall performance than considering alternative options usually employed in MOPSO algorithms. The algorithm and computational results are presented.