Skip to content
Conference Open access

A Unified Ensemble Learning Framework for Multi-Disease Risk Prediction using Standardized Preprocessing and Class Imbalance Handling

Sep 2026 · 2026 10th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC) · pp. 1980-1988 · 0 citations · 27 references

Abstract

Chronic diseases such as cardiovascular disease, stroke, diabetes, asthma, kidney disease and skin cancer carry significant global health burdens and there is a need for predictive systems or models to facilitate early intervention. In this study, we have compared five machine learning algorithms namely Random Forest, Decision Tree, XGBoost, LightGBM and KNN for multi-disease prediction in a database of 254123 patient records using 17 input health indicators and six binary disease targets. Pre-process was carried out, outlier removal, class balanced with SMOTE and feature engineering. The best result was obtained with Random Forest, and across diseases it reached an accuracy of 93.46%, 93.37% and 89.70% for stroke, kidney disease and heart disease, respectively. Competitive performance was achieved by Decision Tree and LightGBM, while KNN continuously lose during cross-validation. These results demonstrate the capability of ensemble approaches to identify more vulnerable populations effectively, while they still face difficulties to accurately predict asthma accuracy 84.30% and diabetes accuracy 84.84%, which can be an avenue recognized for improved feature engineering as well as tailored modeling strategies.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.