Macrosomia Prediction Based on Deep Learning Algorithm

Newborns weighing more than 4000g after birth are collectively called macrosomia. Macrosomia can cause many adverse effects on the mother and the fetus itself. If it is possible to complete the prenatal predictive assessment of a macrosomia before the fetus is born, especially in the early pregnancy, has an important meaning. On the other hand, the combination of traditional clinical medicine and deep learning can reduce the burden of doctors and pregnant women to a certain extent, so as to achieve better diagnosis and intervention. This article applies the method of deep learning to realize the prediction of macrosomia based on pregnancy data. First, using the database provided by the partner with maternal medical indicators, through the database screening and data preprocessing, including digitization, processing of missing values, vectorization and standardization to obtain training datasets. Then we experiment on different machine learning algorithm for comparison, and build a deep neural network dense fully connected layer model, choose the appropriate loss function, activation function and optimization algorithm to achieve higher prediction accuracy. The final goal is to form a macrosomia forecast algorithm to assist the doctors make a diagnose.