<p>The condition known as preeclampsia is a hypertensive disorder that occurs during pregnancy and has serious implications for both the mother and the fetus. Proper management of the condition depends on early detection of preeclampsia to make a correct prognosis. In this study, we classify pre-eclampsia using three datasets: two of which are the public datasets acquired from Mendeley and Kaggle, respectively, while the third is a real-world clinical dataset obtained from a local hospital. Recursive feature elimination, principal component analysis, correlation-based feature selection, and particle swarm optimization were used to select significant features from the predictor variables of the public datasets. To improve the classification performance, several models were created, with an emphasis on ensemble learning methods. Specifically, we propose three models: the alternative classification models include the Soft Decision Fusion Model, which applies soft-voting; the Stacking-Based Classifier, which is an ensemble stacking; and the Hybrid Soft Stacking Model. These models were assessed in detail concerning their quantitative indicators for the AUC-ROC criterion. The performance of our proposed models in the public datasets was an AUC-ROC of more than 95% and in the clinical dataset an even higher 96%. These ensemble methods accurately show that they have effective results in improving the precision and reliability of pre-eclampsia forecasts. With the help of real and public clinical data, the present work presents an effective and ecological approach that can help healthcare professionals make appropriate and timely decisions about the management of pre-eclampsia. In particular, the results of the Hybrid Soft Stacking Model look quite convincing in terms of predictive value, so the model could be considered a useful tool in the clinical context.</p>