A novel RFE-GRU model for diabetes classification using PIMA Indian dataset.

Area of Science:

• Medical Informatics
• Machine Learning
• Data Science

Background:

• Diabetes mellitus is a chronic condition with severe health implications, necessitating early diagnosis.
• Accurate and timely diagnosis of diabetes is crucial for effective management and preventing complications.
• Machine learning offers potential for improving diabetes diagnosis accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a novel machine learning model for early diabetes classification.
• To enhance the predictive accuracy of diabetes diagnosis using the PIMA Indian dataset.
• To compare the performance of the proposed model against established machine learning algorithms.

Main Methods:

• Data preprocessing including mean imputation and normalization of the PIMA Indian dataset.
• Training and evaluation of multiple machine learning models: Random Forest, Logistic Regression, K-Nearest Neighbor, Naïve Bayes, Histogram Gradient Boost, and Gated Recurrent Unit.
• Development of a hybrid Recursive Feature Elimination-Gated Recurrent Unit (RFE-GRU) model for feature selection and classification.

Main Results:

• The proposed RFE-GRU model achieved high performance metrics: 90.50% precision, 90.70% recall, 90.50% F1-score, 90.70% accuracy, and 0.9278 Area Under the Curve (AUC).
• Comparative analysis indicated that the RFE-GRU model outperformed other tested classification models.
• Recursive Feature Elimination effectively identified key predictive features, while the GRU component addressed gradient challenges.

Conclusions:

• The RFE-GRU model presents a robust and effective approach for the early classification of diabetes.
• This hybrid model shows significant promise for improving diabetes diagnosis in clinical settings.
• The findings highlight the potential of advanced machine learning techniques in managing chronic diseases like diabetes.