Evidential Reasoning and Classification in Uncertain Data

Evidential reasoning and classification in uncertain data centres on methods for making decisions when information is incomplete, imprecise or conflicting. At its core lies the representation of uncertainty via belief functions, which allocate support to sets of hypotheses rather than single outcomes. Classification under this paradigm involves constructing basic belief assignments from sensor readings, expert judgements or machine-learning outputs, then fusing these assignments to reach a consensus decision. Key challenges include modelling missing or ambiguous data, resolving conflicts between evidence sources, and ensuring computational tractability in high dimensions. Recent advances have focused on hybridising belief-based fusion with modern learning algorithms—such as decision forests, support vector machines and neural networks—to improve accuracy, robustness and interpretability. Practical applications span environmental monitoring, medical diagnosis, autonomous systems and industrial process control, where the ability to quantify and manage uncertainty directly influences safety and performance.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Evidential Reasoning and Classification in Uncertain Data publication trend

The graph below shows the total number of articles in evidential reasoning and classification in uncertain data across all publications each year (not limited to Nature Index journals).

Technical terms

Dempster-Shafer theory: A mathematical framework for modelling epistemic uncertainty by assigning belief masses over sets of hypotheses.

Basic belief assignment (BBA): A function that quantifies the degree of belief committed exactly to each subset of the hypothesis space.

Belief function: A measure derived from BBAs representing the total support for a proposition, accounting for both direct and indirect evidence.

Evidence fusion: The process of combining multiple BBAs into a single aggregated belief, often via Dempster’s rule or its variants.

Belief entropy: An uncertainty metric based on BBAs that guides attribute selection or decision making by quantifying the dispersion of belief mass.

References

1. Belief Entropy Tree and Random Forest: Learning from Data with Continuous Attributes and Evidential Labels. Entropy (2022).
2. RPREC: A Radar Plot Recognition Algorithm Based on Adaptive Evidence Classification. Applied Sciences (2023).
3. Non-Linear Saturated Multi-Objective Pseudo-Screening Using Support Vector Machine Learning, Pareto Front, and Belief Functions: Improving Wastewater Recycling Quality. Applied Sciences (2024).