This article focuses on the prediction of thyroid diseases and selects 9,172 samples from the Kaggle dataset (2,401 cases with disease and 6,771 cases without disease, including 22 features). First, the Lasso algorithm was used to screen the features to determine 11 features with a high correlation with thyroid diseases, such as TSH, FTI, and T3. After normalization processing by min-max, Four single machine learning models, namely Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GB), and XGBoost, were constructed. An ensemble model was also built with these four models as the base model and Voting Classifier as the meta-model through the Stacking ensemble strategy. After evaluation, the accuracy of the integrated model is 0.95, the recall rate is 0.94, and the F1 score is 0.94, which is superior to the performance of a single model. Research shows that the 11 selected features are closely related to the onset of thyroid diseases, and the constructed integrated model has good predictive performance, providing an effective method for the early prediction of thyroid diseases.