Abstract: In view of the lack of data sets applied to geomagnetic interference classification, the data sets are constructed by processing the geomagnetic interference data, such as data cleaning and label classification, and trained and tested by using the mainstream deep learning algorithm. The experimental results show that under the framework of deep learning, the geomagnetic interference data and undisturbed data can achieve high accuracy classification results, and produce a classification model with certain effect; And its classification effect is related to the establishment of data sample set. A larger and cleaner data sample set will obtain better classification results and more detailed classification ability.