Classification, Evaluation Criteria, and Application Research of the D-S Evidence Reasoning Combination Method

The D-S evidence reasoning combination method has gained significant traction in recent years as a powerful approach for evaluating evidence credibility and facilitating decision-making under uncertainty. This method involves two core computational steps: evidence combination and belief degree calculation. The evidence combination is implemented using Dempster's rule of combination from Dempster-Shafer theory, which mathematically merges multiple evidence sources while handling conflicts through normalization. The belief degree computation employs basic probability assignment (BPA) functions that distribute probability masses to subsets within the frame of discernment, typically implemented through mass function arrays in programming languages like MATLAB or Python. Research demonstrates that the D-S evidence reasoning combination method finds practical applications across diverse domains including fault diagnosis systems, multi-criteria decision support, and risk assessment frameworks. However, ensuring methodological accuracy requires establishing robust evaluation criteria addressing evidence completeness (measured through information entropy), evidence consistency (quantified using conflict measures like K value), and evidence relevance (assessed through correlation coefficients). Algorithm implementations typically involve creating BPA matrices, applying combination rules iteratively, and calculating belief intervals using Plausibility and Belief functions. The classification schemes, evaluation metrics, and practical implementations of the D-S evidence reasoning combination method have been extensively investigated in academic research. Northwest Polytechnical University has contributed significantly through its dissertation "Classification, Evaluation Criteria, and Application Research of D-S Evidence Reasoning Combination Method", which provides comprehensive methodological analysis, code implementation strategies for evidence combination algorithms, and case studies demonstrating practical deployment in engineering systems. This work serves as an essential reference for researchers and practitioners working with uncertainty reasoning and evidence fusion techniques.