Abstract: To improve the accuracy of building settlement prediction, a prediction model (IPSO-SVM) based on improved particle swarm optimization (IPSO) and support vector machine (SVM) is proposed. The IPSO algorithm was used to globally optimize parameters of SVM, so as to improve the generalization performance and prediction accuracy of the model. Based on the existing building settlement sample data, four models were compared and analyzed the prediction effects: IPSO-SVM, PSO-SVM, SVM and BP. The results show that the prediction accuracy of SVM model is significantly improved after the introduction of PSO and IPSO optimization. Among them, IPSO-SVM performs best. The AARD of the prediction results is 8.48%, R² is 0.9369, RMSE is 2.6597, MAPE is 0.0848, and the relative error is controlled below the 11%. Compared with PSO-SVM and SVM model, the prediction accuracy of IPSO-SVM is improved by about 5.4% and 8.4% respectively, which is better than that of BP neural network model. The results show that the IPSO-SVM model has high accuracy and stability in building settlement prediction, which provides a new perspective for the study of building settlement prediction.