<p>Maternity is the crucial period in a woman’s life, and she undergoes many crucial challenges right from the period of conception. Many countries in the world adopt many welfare schemes in order to save the life of the mother and the child, and also to reduce the maternal mortality rate of the country. The main reasons for the maternal mortality include haemorrhage, preeclampsia, gestational diabetes mellitus, etc. The dataset that identifies the risk level of maternal complications is publicly available and includes minimal features for predicting risks. In order to overcome the issue, some real-world data of pregnant women were taken as a reference for the synthetic dataset generation, including more features that are considered in real-world tracking of pregnant women. The generated dataset is then subjected to a stacking classifier with XGBoost and CatBoost algorithms, which produces results with an accuracy of 97.33%. The performance of the state-of-the-art machine learning (ML) algorithms and other boosting algorithms was compared with the stacking classifier. The SHAP algorithm was applied to the prediction model, i.e., the stacking classifier, to get the most contributing factor for the prediction. The dataset seems imbalanced with unequal instances for different class labels. The SMOTE oversampling technique was used to balance the class label instances, and then the stacking classifier was applied to the dataset. The accuracy achieved through this dataset is 96,5%. The other classification metrics, such as precision, recall, and F1-score, are better in the balanced dataset. The work can be extended with more real-life features considered for diagnosis.</p>