Research on Improving Fetal Health Prediction Model Using Optimal Fetal Feature Selection Technique

The phenomena of aging society and declining in birth rate have led to an increased interest in research on predicting fetal health in the field of healthcare. Existing fetal health classification models are actively being studied to improve the models' overall predictive performance. However, predicting fetal health requires a significant number of features and calculations, making rapid and accurate predictions challenging. In this paper, we proposed a method to enhance the accuracy of the fetal health classification models as well as to reduce computational time by selecting the appropriate number of features in the predictive system. Firstly, random resampling method was used to mitigate overfitting caused by data imbalance and normalized the dataset through robust scaling to handle data outliers. Secondly, sequential backward selection of feature algorithm was used to select minimal required features and K-fold cross-validation to ensure the models' accuracy. Using the proposed method, the experimental results achieved 97.2% accuracy, surpassing the results of latest fetal health prediction models studied.