Background:Early prevention of Spontaneous Abortion (SA) is essential for the treatment of recurrent spontaneous abortion. Objective:In this retrospective study, a variety of machine learning methods were used to develop predictive models and diagnose the potential risk of SA.Methods:A total of 663 pregnant women participated in the case-control study, 586 of which were SA patients and 77 were normal parturition women. The research data included 25 features of Traditional Chinese Medicine (TCM) constitution and clinical data related to SA. This work utilized 8 machine learning techniques including logistic regression, gradient boosting decision tree, k-nearest neighbor, classification and r-egression tree, multilayer perceptron, support vector machine, random forest and XG-Boost to predict SA. The performances of the applied models were evaluated by using the method of 10-fold cross-validation and by computing the diagnostic test characteristics, including accuracy, precision, recall, score, and the AUC of ROC curve.Results:The scores of these eight machine learning techniques were all above 97.5%. Among them, gradient boosting decision tree had the best prediction result on SA. The accuracy, precision, recall, score, and the AUC of ROC curve of gradient boosting decision tree were 97.9%, 99%, 98.6%, 98.8%, and 97.3%, respectively. Conclusion:The paper has accurately predicted the risk of SA combined with TCM constitution and clinical data.