A data-driven M2 approach for evidential network structure learning

摘要

The Evidential Network (EN) is a significant model for addressing complex systems under uncertainty, with wide applications in reliability assessment, condition monitoring, and elsewhere. The EN structure is currently often determined by professional experience and expert knowledge, which reduces its objectivity and accuracy. Accordingly, a data-driven model, called M2 approach, is proposed for the EN structure learning, which combines the Maximal Information Coefficient (MIC) and the Additive Noise Model (ANM). More specifically, the MIC is first utilized to generate the initial undirected network structure in line with dependencies of variables, based on which the ANM is then employed to determine the direction of arcs amongst nodes according to information transfer order. A numerical case of three-node EN is applied to validate the feasibility and efficiency of the proposed M2 approach in theory. Further, the experimental results of the turbofan engine degradation dataset, available on the Prognostics Center of Excellence (PCoE) at NASA Ames, illustrate the applicability of the M2 approach in practice.