A Classification Algorithm-Based Hybrid Diabetes Prediction Model.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Computational Biology

Background:

• Diabetes mellitus is a significant global health concern and a leading cause of mortality.
• Early detection and treatment of diabetes are crucial to prevent severe complications.
• Predictive modeling offers a promising approach for identifying individuals at high risk of developing diabetes.

Purpose of the Study:

• To develop and evaluate machine learning models for precise prediction of diabetes likelihood.
• To compare the performance of various classification algorithms in early diabetes detection.

Main Methods:

• Utilized supervised learning algorithms: Random Forest, Support Vector Machine (SVM), Naïve Bayes, and Decision Tree (DT).
• Employed an unsupervised learning algorithm: k-means clustering.
• Experiments were conducted on two distinct datasets: Frankfurt Hospital (Germany) and Pima Indian Diabetes (PIDD) from UCI repository.
• Data was split into training and testing sets to validate model performance.

Main Results:

• On the Frankfurt Hospital dataset, the Random Forest algorithm demonstrated the highest accuracy at 97.6%.
• On the Pima Indian Diabetes dataset, the SVM algorithm achieved the highest accuracy of 83.1%.
• The performance validation through training and testing sets confirmed the reliability of the results.

Conclusions:

• Machine learning classification algorithms are effective tools for early diabetes detection.
• Algorithm performance varies depending on the dataset, highlighting the importance of dataset-specific model selection.
• The developed models show potential for integration into clinical practice to aid in proactive diabetes management.