A diabetes prediction model based on Boruta feature selection and ensemble learning.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare

Background:

• Diabetes mellitus is a prevalent chronic disease and a leading cause of mortality.
• Current treatments are auxiliary, and many patients face complications due to delayed diagnosis or lack of access.
• Early detection and intervention are crucial for managing diabetes and preventing severe health outcomes.

Purpose of the Study:

• To develop an effective early detection method for diabetes.
• To propose a novel diabetes prediction model leveraging advanced machine learning techniques.

Main Methods:

• Utilized Boruta feature selection to identify salient features from a diabetes dataset.
• Employed K-Means++ algorithm for unsupervised data clustering.
• Implemented a stacked ensemble learning method for robust classification.

Main Results:

• The proposed model achieved a high accuracy rate of 98% on the PIMA Indian diabetes dataset.
• Evaluated performance using accuracy, precision, and F1 index, demonstrating strong predictive capabilities.
• The model's effectiveness in diabetes prediction was validated.

Conclusions:

• The developed diabetes prediction model demonstrates superior performance compared to existing models.
• The findings suggest this model is a promising tool for early and accurate diabetes diagnosis.
• Early detection through advanced models can significantly improve patient outcomes and reduce diabetes-related mortality.