Maternal health still poses one of the major global issues, especially in rural settings where medical professionals are few, public clinics are not easily accessible, and transportation is quite inconvenient. These factors greatly increase cases of morbidity and mortality due to maternal and infant health. Different machine learning (ML) and deep learning (DL) methods like Logistic Regressor (LR), Decision Tree (DT), Gradient Boost (GB), K-Nearest Neighbor (KNN), Random Forest (RF), Extreme Gradient Boosting (XGB), Ensemble Learning Stacking (ELS), Ensemble Learning Bagging (ELB), Recurrent Neural Network (RNN), Gated recurrent units (GRU), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Cat-Boost (CB), Gradient Boosting (GB), and Adaptive Synthetic Over-sampling Technique (ADASYN), are used to handle class imbalance within the dataset. In this work, the ADASYN algorithm was used to balance the dataset. The existing methods (Kyzy et al. in Computer Science and Mathematics 1–19, 2024) presented the results by achieving accuracy on the balanced datasets: RF (85.22%), Voting Classifier (VC) (87.19%), and XGB (85.71%). In contrast, the proposed models—KNN, LR, ELS, ELB, LSTM, RNN, GRU, and CNN—have experimented on both balanced (ADASYN algorithm) and imbalanced datasets. Results of the proposed model with 30,000 datasets (10,000 datasets per class) using ADASYN algorithm—LR (57%), XGB (92%), RF (92%), GB (83%), DT (91%), KNN (90%), ELB (92.33%), ELS (92%), CNN (39%), RNN (63%), LSTM (60%), GRU (61%), and CB (90%). ELB had the highest accuracy at 92.33%. Without ADASYN LR (66%), XGB (82%), RF (79%), GB (74%), DT (80%), KNN (68%), ELB (82%), ELS (85%), CNN (37%), RNN (56%), LSTM (53%), GRU (52%), and CB (80%). The highest accuracy ELS came in with 85%. The results demonstrate how the proposed approaches perform significantly better on the ADASYN-generated balanced dataset.