An Interpretable AI-Based Hybrid Model for Accurate AnaemiaPrediction Using Explainable Machine Learning

Abstract

Anaemia is a widespread hematological disorder characterized by reduced haemoglobin concentration, leading to decreased oxygen-carrying capacity of blood and significant health complications if not detected early. Traditional diagnostic methods are often invasive, time-consuming, and lack predictive intelligence for proactive healthcare support. To address these limitations, this paper proposes a transparent anaemia prediction framework using hybrid machine learning integrated with Explainable Artificial Intelligence (XAI).

The system utilizes a structured anaemia dataset containing demographic and clinical parameters such as gender, haemoglobin level, MCV, MCH, and MCHC to classify individuals as anaemic or non-anaemic. Multiple machine learning algorithms including Decision Tree, K-Nearest Neighbors, Support Vector Machine, and Gradient Boosting are implemented and comparatively evaluated.

To enhance predictive performance, a hybrid ensemble model combining Random Forest and XGBoost through a Voting Classifier is proposed, achieving an accuracy of 99.70%. To improve transparency and clinical trust, SHAP and LIME are incorporated to provide interpretable explanations for each prediction by identifying influential features contributing to the model’s decisions.

The complete framework is deployed through a Flask-based web application for real-time prediction and user interaction. Experimental results demonstrate that the proposed model significantly improves prediction accuracy while maintaining interpretability, making it suitable for practical clinical decision support and early anaemia diagnosis.