Optimized classification of diabetes using dynamic waterwheel plant optimization algorithm.

Area of Science:

• Public Health
• Computational Biology
• Machine Learning

Background:

• Diabetes is a prevalent chronic disease requiring accurate classification methods.
• Machine learning algorithms are increasingly used for chronic disease classification.
• Existing methods for diabetes classification need improvement in accuracy and efficiency.

Purpose of the Study:

• To develop a novel and accurate methodology for classifying diabetes.
• To introduce a new feature selection method, the dynamic waterwheel plant algorithm (DWWPA).
• To optimize the K-nearest neighbors (KNN) classification model using DWWPA for improved diabetes categorization.

Main Methods:

• Development of the dynamic waterwheel plant algorithm (DWWPA) for feature selection.
• Utilization of a binary version of DWWPA (bDWWPA) in the feature selection process.
• Optimization of the K-nearest neighbors (KNN) model using the DWWPA feature selection method.
• Comparison of the proposed method against various machine learning models and optimization techniques.
• Statistical validation using Analyses of Variance (ANOVA) and Wilcoxon signed-rank test.

Main Results:

• The proposed DWWPA-optimized KNN method demonstrated superior performance in diabetes classification.
• The method achieved a high classification accuracy of 98.9%.
• Statistical analyses confirmed the significance and superiority of the proposed approach over contemporary methods.

Conclusions:

• The novel DWWPA feature selection and KNN optimization method is highly effective for accurate diabetes classification.
• The proposed approach surpasses existing methods in classifying diabetic disease.
• This study provides a robust and accurate machine learning-based solution for diabetes categorization.