Model-BasedDiagnosiswithMulti-.PDFVIP

Model-Based Diagnosis with Multi-Label Classification Betty Keren, Meir Kalech, and Lior Rokach Information Systems Department, Ben-Gorion University of the Negev, Israel betty.keren1@,kalech,liorrk@bgu.ac.il ABSTRACT ber of components in the system. In case of large scale systems using MDB is impractical due to its Model-Based Diagnosis (MBD) is one of the computational complexity. Unfortunately, the fact leading artificial intelligence approaches that that many of the systems today (e.g. network sys- copes with the diagnosis problem. MBD is tems, robots etc.) have a huge number of compo- known as a hard problem and grows exponen- nents, leads to the infeasibility of MBD approach tially in the size of the system. In this paper, in those systems. we propose a novel approach that combines Many of the previous diagnosis methods pro- MBD with multi-label classification. We pro- pose deterministic reasoning and search algorithms pose to build a classifier that maps symptoms to address the complexity challenge (de Kleer and of the system to possible faults. The major ad- ) Williams, 1987; Williams and Ragno, 2007 . These vantage of this approach is by reducing signif- methods guarantee sound diagnoses;