Abstract: Sunspots occur in the solar photosphere and can make the prediction of solar flares. Aiming at the three classification sunspot data set with unbalanced number of category samples, a dual model integrated algorithm is proposed. This method uses two models, one light and one heavy, to undertake the classification tasks of two categories respectively, and then integrate the classification results of the two to squeeze out the classification results of the third category. Experiments show that this method can reduce the adverse effects of over-fitting and under-fitting of a single model on unbalanced data sets, and solve the problem of class imbalance in sunspot data sets from a new perspective, with an average F₁ score of 0.931.