Newborn Weight Prediction And Interpretation Utilizing Explainable Machine Learning

Newborn weight refers to the weight of a baby at the time of birth. It is typically measured in units such as pounds and ounces or kilograms. The weight of a newborn is an essential metric used by healthcare professionals to assess the health of an infant and its development. This research emphasizes the prediction of newborn weight using Machine Learning (ML) and Explainable Artificial Intelligence (XAI). The proposed approach involves preprocessing a dataset encompassing various parameters of newborns and their mothers during pregnancy. The primary objective is to develop an automated system capable of accurately predicting the weight of a newborn. Employing a range of regression ensemble ML models, namely Bootstrap Aggregating regression (BAGGINGR), Random Forest Regression (RFR), Ensemble Voting Regression (VOTIINGR), Extreme Gradient Boosting Regression (XGBOOSTR), Stacked Generalization Regression (STACKINGR), and Gradient Boosting Decision Trees Regression (GBDTR). This research develops a cross-validation approach, and notably, the GBDT model emerges as the top performer, yielding impressive results with average metrics Mean Squared Error (MSE) of 247.26, Mean Absolute Error (MAE) of 12.29, R-squared (R2) of 0.23, Peak Signal-to-Noise Ratio (PSNR) at 20.99 dB, and Signal-to-Noise Ratio (SNR) at 48.65 dB. The outcomes of the research are interpreted through Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP) in XAI, emphasizing the significance of interpreting these parameters. The proposed research has valuable insights to enhance the long-term health conditions of newborns, reduce mortality rates, and provide crucial support to healthcare professionals.