Complex systems approximate matching approach for large graphs classification optimized by NSGA-II

Complex systems are strongly emerging in various domains from defense/space to network of enterprises. Graph modeling is extensively used to represent these complex systems. The design of these complex systems are increasingly under stringent constraints in design cost and Time To Market (T.T.M.). It is of paramount importance to exploit “design-and-reuse” approaches in building these systems. The reuse can be based on subsystems or systems. Reuse requires identification of graph representations of these subsystems and systems in the large graph representation of complex systems. We propose a combined approach for large graph classification approach based on an approximate matching method and genetic algorithm. The first stage of this method is to perform the comparison on simpler graphs called prime graphs in order to refine the time complexity. The second stage quality of the classification is improved through multiobjective optimization with NSGA II. The values to be optimized are the recognition rate and the confusion rate. Experiment demonstrate the validity of our approach for complex systems.