Machine Learning-Based Classification and Prediction Framework for Early Disease Detection in Healthcare: A Comparative Study

V. Ellappan, S. Tamilarasi, M. Nishanth, S. Vinothkumar, Adam Marks Department of Electronics and Communication Engineering, Mahendra Institute of Technology, Namakkal, Tamil Nadu, India. Department of Information Science, Institute for Higher Education Leadership, Dubai, United Arab Emirates.

Abstract: Predicting a disease in time and accurately is one of the most important challenges in current healthcare. The present study conducts a thorough comparative analysis of the performance of six machine learning (ML) classification algorithms across six publicly available medical datasets: diabetes, heart disease, breast cancer, sepsis, and lung cancer. Researchers proposed an XGBoost-based framework with SHAP (Shapley Additive exPlanations) interpretability, achieving an overall classification accuracy of 95.1%. At the same time, precision, recall, and AUC-ROC were 94.6%, 93.8%, and 0.98, respectively, surpassing all baseline models. All datasets underwent the same extensive preprocessing, including missing-value imputation, SMOTE-based class balancing, and mutual information-based feature selection. Ensemble tree-based techniques do indeed outperform linear approaches on heterogeneous healthcare data, as the results confirm. The most clinically relevant predictors identified in the SHAP analysis are glucose concentration, age, BMI, and troponin level. In this work, a reproducible, explicative, and scalable approach to clinical decision support in resource-limited healthcare settings is given. The study's key findings are clearly outlined, focusing on the superior performance of the comparative analysis of several machine learning algorithms, the adoption of an efficient prediction framework, and its potential applications in healthcare settings where accurate disease diagnosis is crucial.

---

Keywords: Machine Learning (ML); Healthcare Informatics; Disease Prediction; Classification Algorithms; Electronic Health Records; Clinical Decision Support; Tree-Based Techniques.

---

Received on: 12/07/2025Revised on: 17/09/2025Accepted on: 22/11/2025Published on: 07/03/2026

---

---