Objective：To compare the predictive performance of constructing the early prognostic model of acute paraquat（PQ）poisoning between support vector machine（SVM）and logistic regression. Methods：A total of 152 patients with acute PQ poisoning were collected and the clinical results were observed for 2 months. The patients were divided into two groups with a 3∶2 ratio by the random numerical table method. One group with a total of 91 cases was used as a training sample for selecting variables and establishing predictive models. Another group with a total of 61 cases was used as a validation sample to evaluate the predictive effect of the model. SVM and conventional logistic regression was used as the modeling method. Results：The prediction accuracy of the kernel，polynomial，sigmoid kernel and radial basis function nuclear SVM model was 77.92%，74.03%，75.32% and 79.22% respectively，when being tested by the validation group. The results of performance comparison showed that SVM models performed better than logistic regression model；RBF-SVM was the best among all the models with a sensitivity of 87.5% and a specificity of 70.6%. Conclusion：SVM model could preferably integrate all kinds of prognostic information of PQ poisoning patients，and the established model had better prediction ability，providing a new method for predicting the prognosis of patients with PQ poisoning.