Abstract: To study the specific effects of the main chemicals in cigarette smoke on sensory coziness and determine the key factors.42 smoke chemicals in 50 domestic and imported representative tobacco leaves were analyzed.Meanwhile,a comprehensive sensory coziness evaluation was conducted on the 50 samples.Statistical models for the effects of smoke chemicals on sensory coziness were obtained by stepwise regression method.The results showed that the obtained models were statistically highly significant ( pF p <0.05),which signals that all the variables in the models were important for sensory coziness.The obtained models perform satisfactorily in the prediction for test samples,which indicates that the models can be reliably used to predict sensory coziness.The models could provide a scientific foundation for the development and maintenance of tobacco products.