Revolutionizing Diabetes Diagnosis: Machine Learning Techniques Unleashed.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Data Science

Background:

• Diabetes mellitus disrupts glucose metabolism, impacting cellular energy.
• Accurate prediction of diabetes is crucial for timely intervention and management.

Purpose of the Study:

• To predict diabetes occurrence using machine learning algorithms.
• To compare the performance of multiple algorithms on the PIMA diabetes dataset.

Main Methods:

• Utilized machine learning algorithms: Decision Tree, K-Nearest Neighbor, Random Forest, Logistic Regression, and Support Vector Machine.
• Employed Waikato Environment for Knowledge Analysis (WEKA) and Python for experimental execution.
• Evaluated algorithm performance using metrics like precision, recall, F-measure, and ROC AUC.

Main Results:

• Logistic Regression achieved the highest precision: 81% with Python and 80.43% with WEKA.
• Various performance metrics and error analyses (MAE, RMSE) were used to assess model accuracy.
• All tested algorithms showed varying degrees of effectiveness in diabetes prediction.

Conclusions:

• Machine learning, particularly Logistic Regression, shows significant potential for diabetes prediction.
• The study highlights the utility of computational approaches in diagnosing and managing diabetes.