Toward reliable diabetes prediction: Innovations in data engineering and machine learning applications.

Area of Science:

• Medical Informatics
• Computational Biology
• Data Science

Background:

• Diabetes mellitus is a chronic metabolic disorder characterized by hyperglycemia, increasing risks for cardiovascular diseases, nephropathy, and retinopathy.
• Early diagnosis of diabetes is crucial for effective management and prevention of severe complications.
• Machine learning (ML) offers promising tools for developing accurate and efficient diabetes diagnostic models.

Purpose of the Study:

• To develop and evaluate machine learning models for accurate and early diagnosis of diabetes.
• To investigate the impact of data preprocessing techniques, including random oversampling, on model performance.
• To identify the most effective ML algorithms for diabetes prediction across diverse datasets.

Main Methods:

• Implemented a comprehensive data preprocessing pipeline, including random oversampling for handling imbalanced datasets.
• Experimented with four distinct diabetes datasets to assess model generalizability.
• Evaluated multiple machine learning algorithms for their predictive accuracy in diagnosing diabetes.

Main Results:

• Random forest achieved 86% and 98.48% accuracy on Datasets 1 and 2, respectively.
• Extreme gradient boosting and decision tree models reached 99.27% and 100% accuracy on Datasets 3 and 4, respectively.
• The proposed preprocessing methods improved model accuracy by up to 12.15% compared to models without preprocessing.

Conclusions:

• The developed ML models demonstrate high accuracy in diabetes prediction, outperforming existing methods.
• These advanced models can significantly enhance current diabetes screening and diagnostic capabilities.
• The findings support the integration of ML-driven predictions into preventative healthcare strategies to reduce diabetes incidence and associated healthcare costs.