Maternal health risk assessment is essential for ensuring the well-being of both mothers and their developing fetuses during pregnancy. Complications can arise during this period, making accurate risk prediction critical for timely intervention. This study focuses on predicting these specific maternal health risks using advanced boosting techniques and feature selection methods. Specifically, we evaluated various combinations of machine learning algorithms and dimensionality reduction techniques, ultimately identifying Uniform Manifold Approximation and Projection (UMAP) combined with XGBoost (XGB) as the most effective approach. UMAP was selected for its ability to efficiently reduce the dimensionality of the dataset while preserving important structures in the data, thereby improving model performance. XGBoost was chosen due to its reputation for achieving high accuracy in classification tasks, especially when combined with robust feature selection. The performance of the UMAP + XGB model significantly outperforms other combinations, achieving a precision of 98.88%, recall of 98.85%, F1 score of 98.85%, and accuracy of 98.85%. These results underscore the effectiveness of this approach in improving maternal health risk predictions. Furthermore, we compare the performance of the proposed model against current state-of-the-art methods to validate its superiority in terms of predictive accuracy.