Explainable Ensemble Learning For Transparent Anaemia Diagnosis and Clinical Decision Support

Abstract

Anaemia remains a major public health concern in India, particularly among
women and children, where prevalence rates are alarmingly high as reported by NFHS-5.
Despite national initiatives such as Anaemia, challenges like delayed diagnosis, limited
screening access, and nutritional deficiencies continue to increase disease burden and healthcare
costs. Traditional diagnostic methods rely on laboratory tests and clinician interpretation of
haemoglobin levels and patient history, which are often time-consuming, subjective, and not
suitable for large-scale screening. To address these limitations, this study proposes a machine
learning–based anaemia prediction system integrated with Explainable AI (XAI) to ensure both
accuracy and interpretability. The system utilizes algorithms such as SVM, Decision Tree, KNN,
Gradient Boosting, along with ensemble methods like Random Forest and XGBoost to predict
anaemia from patient data. Model performance is further improved using voting techniques to
enhance prediction accuracy. In addition, explain ability techniques such as SHAP and LIME are
incorporated to provide transparent insights into model decisions by highlighting the contribution
of key features like haemoglobin levels, age, and gender. This improves clinical trust and allows
medical experts to validate predictions effectively. The proposed system enables early detection,
scalable screening, and reliable decision support, making it suitable for real-world healthcare
applications