Development of an Integrated Evidential Reasoning Approach Using Both Belief and Disbelief Structures with Validation in Multicriteria Assessment of Organizational Innovation Management

Project: Research

View graph of relations


Forming sound and reliable evaluations under various judgmental uncertainties has become a challenge to the classic multiple criteria decision making (MCDM) theory, and calls for the development of scientific approaches that are capable of handling various judgmental uncertainties in a way that is rational, reliable, flexible, and transparent. The evidential reasoning (ER) approach provides a novel, rigorous, and consistent way of modelling uncertainties in MCDM, and has gained increasing attention and extensive applications in a wide variety of areas in recent years. In the ER approach, the subjective judgments, uncertainties, and numerical data on quantitative decision criteria decision makers (DMs) are all modeled with belief structures that are viewed as evidence and aggregated with the Dempster-Shafer (D-S) theory of evidence.The use of the ER approcah for solving complex MCDM problems is still subject to limitations, and has room for improvement. First, judgmental uncertainties occasionally need to be modeled using belief and disbelief structures simultaneously, whereas the existing ER approach can only model them with belief structures. Second, the ER approach treats missing data as outright ignorance without addressing proper approaches to learn them. This calls for decisions to be made under great uncertainty, increasing the risks of making an unreliable decision. Third, the ER approach does not address the issue of how the weights of the decision criteria are determined, and assumes them to be known or having been specified by the DM. Fourth, the ER approach needs to transform numerical data on quantitative decision criteria into belief structures for aggregation. This requires extra computational efforts, and no approach or algorithm has been proposed for aggregating them directly without the need for data transformation. Fifth, complex MCDM problems may occasionally contain ordinal data. The ER approach does not address the issue of modeling ordinal data and cannot meet the needs of handling ordinal data. Finally, the existing ER approach compares and ranks decision alternatives in light of their expected utilities, whose calculation requires the utilities on each assessment grade of the DM to be known or specified. No approach or model has been proposed to estimate these data without specifying them subjectively to produce a full rank of decision alternatives.Organizational innovation management (OIM) assessment is a typical complex MCDM problem with both quantitative and qualitative assessment criteria. OIM requires a large number of subjective judgments from DMs or experts and inevitably involves various judgmental uncertainties, such as imcomplete judgments, outright ignorance, and so on. Therefore, OIM assessment often has to be made under uncertainties.This project is proposed to address the above-mentioned limitations of the existing ER approach and the need for an effective OIM assessment system. A new integrated ER approach for complex MCDA using both belief and disbelief structures is developed. An approach for determining the missing data is proposed. Methods for assessing the weights of decision criteria from decision data are suggested. Algorithms for aggregating belief and disbelief structures and numerical data without the need for data transformation are derived. Procedures for modeling ordinal data under the framework of D-S theory of evidence are investigated. A data envelopment analysis (DEA)-based convex quadratic programming model for estimating the utilities of assessment grades is provided. The proposed new ER approaches, models, and algorithms are applied in developing an OIM assessment system with the validation of collaborating companies. The project results can contribute to the Hong Kong research base, and to industrial applications in MCDM and OIM areas.


Project number9041660
Grant typeGRF
Effective start/end date1/09/1116/02/16