Abstract: Aiming at the problem of insufficient precision in emotional neural network (ENN) due to the lack of dynamic response capability when handling complex prediction problems, a self-organizing recurrent emotional neural network based on cell competition and hormone regulation (CHSORENN) is proposed. First, to enhance the learning ability and dynamic characteristics, the hormone regulation and a recurrent structure are integrated into the ENN. Through feedback loops and the hormone regulation mechanism, the historical information from the previous time step is returned and used to regulate the input. Second, a structural adjustment mechanism based on cell competition is proposed to enhance the interpretability and generalization ability of structural adjustments. An adjustment set is used to determine the range of neurons that require adjustment, and then precise pruning is achieved through neuron significance analysis to determine the optimal network structure. Finally, an improved particle swarm optimization based on subgroup leader and Gaussian mutation (SLGPSO) is used to optimize network parameters, balancing exploration and exploitation, avoiding local optima, and improving the network prediction precision. Experimental results show that the proposed CHSORENN has good dynamic characteristics and can achieve significant improvements in both network structure compactness and convergence precision.