Aiming at the problem of wine identification, the odor information of 7 kinds of wine was collected through the electronic nose, the LightGBM algorithm was used to learn the odor characteristics of the wine, and the TPE hyperparameter optimization algorithm is used to adaptively optimize the HyperGB parameter of the LightGBM algorithm. Verification is an indicator to evaluate the performance of the model. The experimental results showed that the discrimination model established by LightGBM had a 96.62% accuracy rate for wine samples, which was superior to traditional support vector machines, random forests, and neural networks. It verifies the superiority of LightGBM in wine variety identification and provides wine identification a fast, reliable and effective analysis method is also suggested, and more excellent algorithms can be introduced into the field of wine smell data mining machines.