Diagnosis Rule Mining of Airborne Avionics Using Formal Concept Analysis

A diagnosis rule mining approach based on formal concept analysis is proposed to discover diagnosis knowledge in airborne avionics. Diagnosis formal context is defined to organize fault samples in 0-1 table. Concept lattices are constructed from the diagnosis formal context, using a parallel Next Closure algorithm applying MapReduce framework. Knowledge reduction in diagnosis formal context is discussed to get reduced attribute sets. The discernibility matrix and Boolean function defined in a rule acquisition perspective are used to calculate all the reduced attribute sets. Compact diagnosis rules can then be derived from the reduced concept lattices. The proposed diagnosis rule mining approach is demonstrated with the historical diagnostic dataset of some aviation radar system.