Decision-making could be defined as the process to choose a suitable decision among a set of possible alternatives in a given activity. In this field, Fuzzy Cognitive Maps (FCMs) have gained significant attention for their ability to model complex systems through causal relationships between concepts. While FCMs are transparent and adaptable models, challenges arise in balancing stability and accuracy, especially in classification tasks. This paper presents an FCM-based classifier using the relative activation value (RAV) methodology to enhance performance in classification scenarios. The model adopts a class-per-output topology and employs backpropagation with quasi-nonlinear inference to optimize learning. Additionally, a novel loss function is introduced to improve convergence and classification accuracy. Experimental results demonstrate that the proposed classifier achieves a notable \(7\text{\%}\) improvement in accuracy compared to a similar classifier optimized with a population-based method, reaching an average classification accuracy of \(82\text{\%}\) . This study underscores the potential of FCMs for transparent and efficient classification, with further enhancements possible through parameter optimization.