Learning of Sparse Fuzzy Cognitive Maps Using Evolutionary Algorithm with Lasso Initialization

Fuzzy cognitive maps (FCMs), characterized by a great deal of abstraction, flexibility, adaptability, and fuzzy reasoning, are widely used tools for modeling dynamic systems and decision support systems. Research on the problem of finding sparse FCMs from observed data is outstanding. Evolutionary algorithms (EAs) play a key role in learning FCMs from time series without expert knowledge. In this paper, we first involve sparsity penalty in the objective function optimized by EAs. To improve the performance of EAs, we develop an effective initialization operator based on the Lasso, a convex optimization approach. Comparative experiments on synthetic data with varying sizes and densities compared with other state-of-the-art methods demonstrate the effectiveness of the proposed approach. Moreover, the proposed initialization operator is able to promote to performance of EAs in learning sparse FCMs from time series.