Maternal health issues pose significant risks to pregnant women, often leading to complications arising from conditions such as diabetes or abnormal glucose levels, depression, hypertension, anxiety, and other disorders. Early identification and monitoring of risk factors are crucial to minimizing such complications. This study leverages real-world data to identify and predict maternal health risks using a machine learning (ML)-based system designed to forecast the likelihood of maternal illness. Multiple ML models were integrated into the system to enhance the accuracy of the prediction. A dataset collected from maternity hospitals and clinics was subjected to four different training and testing scenarios. Exploratory data analysis revealed hypertension, hypotension, and diabetes as the primary contributors to complications. The proposed methodology introduced a novel approach to addressing high-risk factors, emphasizing class-specific performance to better distinguish between low, medium, and high-risk cases. Among the models, using the random forest classifier, we have achieved exceptional performance, delivering a success rate of 91% in high-risk class predictions and an overall assessment score of 0.915 across all classes. Class-wise performance evaluation emerged as a key strength of this approach, enhancing its ability to accurately identify and prioritize risks. By enabling early detection of high-risk pregnancies, this predictive system holds promise for timely intervention and treatment, potentially reducing pregnancy-related complications and saving lives. The findings underscore the importance of advanced predictive tools in addressing maternal health concerns and fostering broader awareness within the healthcare community.