Early Predicting Congenital Fetus Malformations Based on Decision Trees Algorithm

The paper presents an effective system for early prediction of the probability of a deformed fetus birth using a decision trees algorithm. A number of cases were collected between 2020 and 2023 from the Children’s Hospital in Damascus, Syria, where the total number was 152 cases of healthy newborns and premature babies and others with deformities. 12 attributes were extracted and organized in database; they were reduction to 10 important features. Features are (Father’s age, mother’s age, smoking during pregnancy, Having Medications during pregnancy, diseases and infections during pregnancy, genetic factors, family history, kindred between the Couple, exposure to rays during pregnancy, previous miscarriages or deaths). Data was cleaned and prepared to suitable form for database of deformed fetus Characteristics, then they encoded and scaled to be a good input of the designed neural network. NN was designed according to the following parameters, which are criterion = “entropy”, splitter = “best”, max_features = “sqrt” and the NN output was considered yes or no i.e. either the fetus is deformed or not. NN was trained depending on the decision trees algorithm using the fitting function. Data was split into two groups, 80% for training set and 20% for testing set. NN was trained and tested on new samples, and the classification accuracy rate reached to 0.93%. Experiments results confirm that our system is capable of predicting early and with a high accuracy and can spare many families suffering for many years. Therefore, it can be relied upon in many medical centers to be an effective and quick tool in the doctor’s hands, furthermore it an important indicator with a small error rate, addition to it saves a lot of material, financial and moral costs.