New methods based on a genetic algorithm back propagation (GABP) neural network for predicting the cigarette ventilation rate

Abstract The ventilation rate of cigarettes is an important indicator that affects the suction resistance of cigarettes. This paper used multiple linear regression networks (MLR), backpropagation neural networks (BPNN), and genetic algorithm-optimized backpropagation (GABP) to construct a model for the prediction of cigarette ventilation rate. The model results indicated that the total ventilation rate was significantly positively correlated with unit weight(P<0.01), circumference, hardness, filter permeability, and suction resistance. The results showed that the MLR models' (RMSE = 0.651, R2 = 0.841) and the BPNN models' (RMSE = 0.640, R 2 = 0.847) prediction ability were limited. Optimization by genetic algorithm (GA), GABP models were generated and exhibited better prediction performance(RMSE=0.606, R2=0.873). The results indicated that the GABP model has the highest accuracy in the prediction of predicting ventilation rate and can accurately predict cigarette ventilation rate. This method can provide theoretical guidance and technical support for the stability study of the ventilation rate of the unit.