Data Normalization in Decision Making Processes

• Autor(a) principal: Vafaei, Nazanin
• Data de Publicação: 2021
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/131054
• topic: Multi Criteria Decision Making; MCDM; Normalization; Dynamic MCDM; Data Fusion; Aggregation; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

With the fast-growing of data-rich systems, dealing with complex decision problems is unavoidable. Normalization is a crucial step in most multi criteria decision making (MCDM) models, to produce comparable and dimensionless data from heterogeneous data. Further, MCDM requires data to be numerical and comparable to be aggregated into a single score per alternative, thus providing their ranking. Several normalization techniques are available, but their performance depends on a number of characteristics of the problem at hand i.e., different normalization techniques may provide different rankings for alternatives. Therefore, it is a challenge to select a suitable normalization technique to represent an appropriate mapping from source data to a common scale. There are some attempts in the literature to address the subject of normalization in MCDM, but there is still a lack of assessment frameworks for evaluating normalization techniques. Hence, the main contribution and objective of this study is to develop an assessment framework for analysing the effects of normalization techniques on ranking of alternatives in MCDM methods and recommend the most appropriate technique for specific decision problems. The proposed assessment framework consists of four steps: (i) determining data types; (ii) chose potential candidate normalization techniques; (iii) analysis and evaluation of techniques; and (iv) selection of the best normalization technique. To validate the efficiency and robustness of the proposed framework, six normalization techniques (Max, Max-Min, Sum, Vector, Logarithmic, and Fuzzification) are selected from linear, semi-linear, and non-linear categories, and tested with four well known MCDM methods (TOPSIS, SAW, AHP, and ELECTRE), from scoring, comparative, and ranking methods. Designing the proposed assessment framework led to a conceptual model allowing an automatic decision-making process, besides recommending the most appropriate normalization technique for MCDM problems. Furthermore, the role of normalization techniques for dynamic multi criteria decision making (DMCDM) in collaborative networks is explored, specifically related to problems of selection of suppliers, business partners, resources, etc. To validate and test the utility and applicability of the assessment framework, a number of case studies are discussed and benchmarking and testimonies from experts are used. Also, an evaluation by the research community of the work developed is presented. The validation process demonstrated that the proposed assessment framework increases the accuracy of results in MCDM decision problems.