Establish a prediction model for sensory quality of tobacco leaf alcoholization, which is used to predict sensory changes in the process of tobacco leaf alcoholization process, breaking through the traditional empirical method to guide tobacco leaf inventory turnover and formulation use. In this paper, 18 alcoholic sensory scores of the 18 different samples in 4 warehouses were taken as initial samples at intervals of 6 months, totally 36 months. Firstly, Through factor analysis, the tobacco alcoholization level is divided, and then to perform BP neural network test and prediction,the results show that this model is feasible and effective for predicting the sensory quality of tobacco leaf alcoholization, and Mean Squared Error(MSE) is 1.00E 06. It is predicted that the degree of alcoholification of tobacco leaves in the 42th month can only be seen as a failure of Liangshan Huili C3F。