Dual-Perspective XAI for Preeclampsia Risk: Shap and Lime Analysis

Pre-eclampsia is a dangerous hypertensive condition that can lead to serious health problems for both mother and foetus during pregnancy. Therefore, it is of great importance to determine the risk at an early stage and to predict possible adverse outcomes. In this study, the effects of clinical parameters associated with pre-eclampsia on classification models were evaluated by analyzing two different data sets. In the first dataset, multiple classification was performed with low, medium and high-risk levels, while in the second dataset, adverse pregnancy outcomes were binary classified as ‘yes’ or ‘no’. The performance of the models created with Random Forest, XGBoost and Logistic Regression algorithms for both data sets was measured by accuracy, precision, recall, F1 score LIME and SHAP analysis. Explanations made with the LIME and SHAP methods revealed that some common clinical markers were prominent in both datasets. These findings suggest that certain biomarkers may be important indicators for predicting both pre-eclampsia risk and pregnancy outcomes.