HYBRID DECISION-MAKING MODELS FOR MULTI-CRITERIA OPTIMIZATION PROBLEMS

Abstract

Most of the real-world problems are dependent on a set of criteria rather than a single criterion, which are usually tangible and their corresponding information is merely impossible to get in quantitative form. Hence, this increases difficulties for a method to handle such complexities and provide a feasible solution. If the Decision Makers (DMs) are involved too, it can worsen the decision output. Thus, the development of efficient Hybrid Decision-making (HDM) models is a need of the hour that can provide the optimal solutions to the DMs. The main focus of this thesis is to develop efficient Hybrid Multi-Criteria Decision Making (HMCDM) models for handling fuzzy environment, sustainability issues, rankability issues, multi-dimension data and incomplete data. The HMCDM models are developed by integrating the different Multi-Criteria Decision Making (MCDM) methods, followed by the sensitivity analysis and correlation analysis for validation purposes in different chapters. The application of the proposed integrated model is demonstrated on a set of three problems taken from different Indian industries, including the pulp and papermaking industry, packaging industry, and paper mills. Four types of HMCDM models are considered (I) integration of MCDM methods with MCDM methods, (II) integration of MCDM with Fuzzy Set Theory, (III) integration of MCDM with MODM method, and (IV) integration of MCDM with a sports ranking method (D-matrix). The entire work done for this research is organized into six chapters of the thesis. Chapter 1 defines multicriteria decision-making in the context of decision analysis in general and draws the distinction between single and multicriteria decision-making. It provides a classification of MCDM methods into those based on multi-attribute utility theory, outranking methods and hierarchical approaches. It further provides elementary decisions and distinguishes between multi-objective decision making and multi-attribute decision making and illustrates their historical development along with a brief literature review of some selected weighting methods. An analysis of published literature provides some insights into recent trends. Chapter two provides a brief summary of the main MADM and MODM methods used in this thesis. This includes mathematical formulations and brief descriptions of the steps required for application of a method.Chapters 3 to 5 contain the main contributions of the thesis. Each of them develops one or two HMCDM methods and applies them to a particular industrial application. Third chapter deal with the selection problem while considering the fuzziness in the DMs’ opinions and information collected. In this chapter, two different HMCDM models are proposed for both the selection problems of the pulp and papermaking industry and the packaging industry in India. The fourth chapter draws the attention towards another ranking problem in MCDM. Here, MCDM method is integrated with a minimization model to overcome the drawback of MCDM methods of providing conflicting rankings. An example of Pulp and paper industry in context to India is considered to demonstrate the proposed framework. Here, the performance of twenty-two Indian paper mills is measured by seven MCDM methods and finally, an optimal aggregated rank is acquired. The fifth chapter covers the issue of incomplete information in the MCDM problem by discussing the various imputation in the past and then proposes a method Modified D-Matrix (MoDM) method. This method can handle biased and bias free evaluations in both result-separating and result-merging contexts. The proposed method is applied to the supplier selection data, and the results were found to be in accordance with experts. Finally, the major contributions of this thesis along with concluding remarks in theoretical and application facets are presented in Chapter six. The results obtained in all the chapters indicate that the performance of MCDM can be improved significantly by integrating it with either classical MCDM or artificial intelligence techniques like Fuzzy set theory, Genetic Algorithm.