Traditional cloud classification methods exhibit limitations such as subjectivity and low efficiency and accuracy. To address these issues, a cloud classification prediction model based on the light gradient boosting machine (LightGBM) was proposed. Firstly, feature variables, including cloud top height, cloud bottom height, cloud layer thickness, average reflectivity factor, liquid water content, and duration obtained through millimeter-wave radar were utilized. A dataset was then constructed by combining these features with classification labels to meet the requirements of the model. This dataset was subsequently used to build a classification model that categorizes clouds into seven types: St, Sc, Cu, As, Ac, Cs, and Cc. The experimental results demonstrate that the model achieves an accuracy of 94.70%, precision of 94.68%, recall of 94.97%, and F₁ of 94.65%. These results indicate superior classification performance compared to other models. Therefore, the constructed LightGBM model shows significant effectiveness in cloud classification and recognition, exhibits strong applicability, and holds promising prospects for the automation of cloud recognition services.