A Multi-Objective Hybrid Optimization Algorithm for Project Selection Problem
Babak Amiri · 2012
Choosing a portfolio of projects that meets an organization's objectives without exceeding available capital resources is recognized as a critical issue for which the decision maker takes several aspects into consideration. As most of these aspects may be conflicting, the problem can be considered as a multi-objective one. Correspondingly, in this study a special Multi-Objective Evolutionary Algorithm (MOEA) based on harmony search algorithm (HAS) is designed to solve the project selection problem. The original HAS often converges to local optima which is a disadvantage with this method. To avoid this shortcoming the HAS was combined with a Chaotic Local Search (CLS). In the proposed algorithm an external repository considered to save nondominated solutions found during the search process and a fuzzy clustering technique was used to control the size of the repository. The experiment results show the capability of the proposed multi-objective algorithm in project selection problem. Selection of right sets of projects is considerably critical for organizations to successfully achieve their competitive advantages and corporate strategies. Due to limited resources and dynamic changes in business environment, this kind of selection is quite challenging for organizations. Beside one hundred selection tools and techniques, academics and practitioners have studied and recommended complex selection methodologies to facilitate the selection of right projects. Some of these methodologies have limitations in various aspects. Among the available useful approaches, optimization techniques (i.e. goal programming, multi-criteria decisionmaking, etc.) are the most fundamental quantitative tools for project portfolio selection which can address a high percentage of desired aspects. Doerner et al. [1] proposed a multi-objective Pareto Ant Colony Optimization which introduces Pareto Ant Colony Optimization as an especially effective meta-heuristic for solving the portfolio selection problem and considers the model introduced by Stummer and Heidenberger [2]. Badri, Davis, and Davis [3] introduced a comprehensive mixed 0–1 goal programming model for project selection in health service institutions. Mukherjee and Bera [4] provided an application of goal programming project selection decision with a case study from the Indian coal mining industry. They instituted a framework for incorporating ratings from experts to compute goal weights with normalization of deviational variables and stochastic demands. Santhanam and Kyparisis [5], proposed a multiple criteria decision model for information system project selection using a nonlinear 0–1 goal programming model that took advantage of hardware and software sharing for IS applications. Santhanam and Kyparisis [6], also discussed a nonlinear 0–1 decision model for interdependent information system project selection formulating benefit, resource, and technical interdependencies among candidate projects. Dey [7] proposed a project evaluation and selection by developing a decision support system. This DSS – applied in an Indian oil pipeline project-analysed project with respect to market, technicalities, and social and environmental impact in an integrated framework using analytic hierarchy process and a multiple-attribute decision-making technique. Gabriel, Kumar, Ordonez, and Nasserian [8] proposed a unique multi-objective project selection model with probability distributions to describe costs and incorporating Monte Carlo simulation and AHP. Medaglia, Graves, and Ringuest [9] described an evolutionary approach for project selection problems with partially funded projects, multiple (stochastic) objectives, interdependencies in the objectives, and a linear structure for resource constraints. Mavrotas, Diakoulaki, and Kourentzis [10] proposed a two-phase method: (1) projects are ranked by a multi-criteria approach. (2) The preorder of projects is used in an integer programming model to derive the final selection. Carlsson, Fuller, Heikkila, and Majlender [11] presented a fuzzy mixed-integer programming model for R&D portfolio selection problem. Huang [12] incorporated random fuzzy uncertainty into project selection by integrating genetic algorithm with random fuzzy simulation. According to the literature review, meta-heuristic approaches in solving project selection problems have a diminutive role in attracting researchers’ attention. In this study, our contribution is to propose a hybrid multiobjective algorithm based on HSA and CLS. The harmony search algorithm was developed by Geem et al [13] and has successfully been applied to various combinatorial optimization problems [14-16]. It applies the musical process of searching for a perfect state of harmony. Musical harmony is analogous to the optimization solution