Wavelet fuzzy cognitive maps

Fuzzy cognitive maps (FCMs) are fuzzy influence graphs which consist of concepts and weighted edges. Various transfer functions have been applied in modelling and simulating the dynamic system of FCMs. In FCMs, transfer function is used to bound the expression level of nodes to a certain range. Therefore, in this paper, we first use wavelet transfer function, and then combine it with FCMs to form wavelet FCMs (WFCM). The wavelet function is a kind of local functions that has limited duration and an average value of zero. Then, we conduct comprehensive analyses over existing transfer functions using synthetic data, real data and pattern classification problems. Finally, according to analysis, a new method involving the selection of transfer functions in the optimization process for pattern classification problems is proposed. The experimental results demonstrate the effectiveness of the proposed method. Still, findings show how the existing functions offer different capacities to deal with both problems.