Abstract: In learning of fuzzy cognitive maps (FCMs), traditional learning algorithms only minimize the data error of the directed arcs between every two concepts, which may lead to the low accuracy of the fitting model. The idea of multi-objective optimization is introduced for designing the learning algorithm of FCMs. The multi-objective model for learning of FCMs is proposed with the objectives of minimizing the data error of the directed arcs and minimizing the proportion of the data error. The proposed multi-objective model can reduce the dependence on the weights of the learning algorithms. A multi-objective evolutionary algorithm based on coordination transformation, termed as multi-objective optimization evolutionary algorithms based on coordinate transformation (MOEA/CT), is used to solve the multiobjective model for learning of FCMs efficiently. The approach of parameter setting for MOEA/CT and computational complexity are studied. Experimental results show that the learning of FCMs by using the proposed MOEA/CT can effectively decrease the data error and the model error and reflect the causal relationship between concepts accurately.