The BP neural network was combined with the genetic algorithm to optimize the amount of spice added， and the prediction model was established to provide the optimal ratio of the flavor of the mutton soup and the removal of the muttony odour. On the basis of single factor experiment， the sensory score of mutton soup was used as the evaluation index to carry out Box-Benhnken design （BBD） to investigate the effect of three kinds of spices on the sensory score of the stewed mutton soup samples. Finally， the results of BBD experiment were taken as the initial population. The addition amount of thyme， ginger and pepper was taken as the input value， and the sensory score of stewed mutton soup was taken as the output value. The BP neural network model was used to debug the function fitness. The BP neural network generalization ability is verified using a data set other than the BBD data. The genetic algorithm was used to predict the global optimal combination of the three flavors added in the experimental level. Analysis of variance（ANOVA） showed that the addition of thyme and ginger had a significant effect on the sensory score of mutton soup （P<0.05）. The correlation coefficients of BP neural network training， testing， and verification data were 0.9914， 1 and 0.9849， respectively， indicating that the BP neural network prediction model has good accuracy and can be used to predict the sensory score of stewed mutton soup. The fitness value of the BP neural network generalization ability verification experiment and the predicted value are R2=0.9922， which indicates that the established BP neural network model has a good generalization ability and can accurately predict the data set not in the training set. The results of using genetic algorithm for optimization were 0.54% thyme， 0.48% ginger， 0.16% pepper. The experimental value of the final sensory score was 20， and the error of the model predicted value was only 2.7%. The above results indicate BP neural network combined with genetic algorithm is a method with better accuracy to optimize the proportion of spices added.