Data Augmentation Using SMOTE Technique: Application for Classification of Patient Care

Abstract

In modern healthcare, machine learning (ML) plays an essential role in improving patient care and classification processes. Unlike traditional approaches that rely on manual diagnosis and limited data analysis, ML algorithms can efficiently process large datasets, detect patterns, and generate accurate predictions. Despite its potential, one of the main challenges in applying ML in healthcare is the presence of small and unbalanced data sets that limit the performance of the model. To address this, the Synthetic Minority Oversampling Technique (SMOTE) has emerged as an effective solution by generating synthetic data to balance class distributions. In this study, we evaluated the performance of several ML models, Extra Trees (ET), Random Forest (RF), Gradient Boosting Classifier (GBC), Light Gradient Boosting Machine (LIGHTGBM) and Extreme Gradient Boosting (XGBoost) to assess their predictive accuracy and generalization capability in patient care classification. The results revealed significant improvements for all models when trained in balanced datasets, with the ET model achieving an accuracy of 80%, while performing poorly in imbalanced data.