Background: Preeclampsia affects pregnant women, resulting in stroke, organ failure, and other health problems like seizures. The effect of preeclampsia is most pronounced in developing countries and it affects about 4% of pregnant women causing several illnesses and even death. The key to solving the problem of preeclampsia is its early detection and use of machine learning algorithms that can take various demographic features, biochemical markers, or biophysical features, select important features and find hidden patterns that point to preeclampsia. Objective: The objective of this research is to develop a machine-learning framework to detect Preeclampsia in pregnant women. Method: This research develops a model to detect preeclampsia using principal component analysis (PCA) as a feature selection, k-means as an outlier detection, a combination of SMOTE oversampling, random under sampling and the decision tree (DT) to classify and predict the risk of preeclampsia among pregnant women. The data was obtained from the University of Abuja Teaching Hospital, Abuja, Nigeria. Result: Findings revealed that the combination of the PCA, SMOTE and random undersampling and DT outcome resulted in the best accuracy of 96.8% which is better than the accuracy of existing work (92.1%). Furthermore, the reliability of the model was measured and tested using Bayesian Probability. Conclusion: The developed model can be helpful to Health care providers in checking preeclampsia among women with high blood pressure during their second antenatal visits.