Cardiotocograph is the most commonly used fetal health diagnosis technology. Because of subjective factors and low classification accuracy in traditional manual analysis of Cardiotocograph data, this paper proposes an improved Stacking classification model based on Principal Component Analysis dimension reduction. Firstly, based on the classification effect of learners and the differences among different learners, K-Nearest Neighbor, Gradient Boosting Decision Tree and Support Vector Machines are selected as base learners, and Logistic Regression is used as meta-learner, to form a two-layer Stacking model. In addition, Principal Component Analysis is used to reduce the dimension of high-dimensional data to remove noise and unimportant features. The results show that the Recall and Precision of this model reach 99.3%, which can effectively assist doctors in diagnosis and realize rapid and accurate diagnosis and prediction of fetal health.