The work aims to improve the cutting quality of the cigarette end face, so as to optimize the dynamic balance of the ledger. The dynamic balance function of the ledger was established, with the design angles of two counterweights as the design variables, and the root mean squares (RMS) of the inertia force and torque as the optimal targets. The prediction model for dynamic balance of the ledger was established based on BP neural network, and then a genetic algorithm (GA) was used to optimize such neural network, in order to improve its prediction accuracy and obtain the optimal value. The results showed that the dynamic balance had significantly been improved after the optimization, in which the RMS of the inertia force was decreased by 37.16% and the RMS of the inertia torque was decreased by 8.33%. With the neural network and GA, the dynamic balance of the ledger is optimized in a better way, and the stability of the ledger is improved.