Application of multi-layer deep learning network for prediction of maternal health

In contemporary research, Artificial Intelligence (AI) and Machine Learning (ML) are vital for predicting pregnancy-related issues by analyzing extensive datasets to create synthetic patterns for personalized assessments. This study focuses on maternal health during pregnancy utilizing a deep learning model, particularly a type of ML that employs multi-layered Artificial Neural Networks (ANN) to discern patterns in data. Two robust linear sequential models based on dense networks were developed using Python, tested with datasets from open-source repositories. The models utilized fifteen variables, including fourteen inputs and one output, with birth weight as the outcome variable. The foundational model consists of five dense layers, while the advanced model includes two additional layers, totaling seven. Model performance was assessed through precision, accuracy, F1-score, and recall rate, with data split into 80% for training and 20% for testing. The basic model trained over 100 epochs with a batch size of 16 recorded an F1-score of 86.22%. In contrast, the advanced Dense CNN linear Sequential Maternal-Health (DCNN-SMH) model achieved a higher F1-score of 91.34% and an accuracy rate of 95% for both prediction and classification, outperforming the base model, which had an accuracy of 93%. The study concludes that advanced dense network models yield superior accuracy compared to base neural networks.