Currently, the replacement of tobacco formulation materials is carried out through formulation experiments and experts tasting, which is problematic due to the multitude of material options and complex combinations. It has two weaknesses, a high number of ineffective experiments and significant human resource investment. This paper propose a method based on the Classification Base of Association (CBA) for mining historical formulation material usage rules, which has constructed a scoring recommendation model based on the importance and experts’ attention to material attributes. Experiments on the Yunnan Hongta Group’s last 5 years data demonstrate the CBA tobacco formulation recommendation technology that achieves the capability of tobacco formulation recommendation, with a model accuracy rate of 92%, reducing the material search space and solving problems such as reliance on experience and poor timeliness inherent in human-centric tobacco formulation substitution models.