Abstract: In order to improve the smoking quality evaluation efficiency, Spearman correlation analysis and partial correlation analysis were used to determine the seven chemical components which affected the smoking quality of flue-cured tobacco. The artificial neural network between seven chemical components and smoking quality was established using BP artificial neural network. Finally, the artificial neural network was used to predict the smoking quality of different flue-cured tobacco samples. The results showed that the use of artificial neural network to evaluate the smoking quality of flue-cured tobacco had a good correlation with the artificial method and a low error. Flue-cured tobacco smoking quality evaluation software designed in C++ language can facilitate the training of tobacco smoking quality model and the evaluation of smoking quality, and it had good practicability.

Key words: artificial neural network, correlation analysis, tobacco smoking, evaluation model