Based on the classification and recognition algorithm of machine learning, this paper analyzes and researches the odor drift data set. First of all, data visualization is used to effectively master the data distribution law, coherence, outlier noise points and other information of the data set. According to the situation, the data is normalized and dimensionality reduction preprocessing, and the training set and test set are divided. KNN model, decision tree model, random forest classifier model and MLP multi-layer perceptron model were used to test and compare the data sets. The test results show that the performance of random forest model for odor drift data classification is relatively good, up to 95%, which can be used in practice.