Diagnosis of Chronic Kidney Disease Based on Support Vector Machine by Feature Selection Methods

Area of Science:

• Nephrology
• Medical Informatics
• Data Science

Background:

• Chronic Kidney Disease (CKD) is a progressive condition necessitating early detection and treatment to reduce mortality.
• Machine learning (ML) offers high-accuracy classification for medical diagnosis.
• Effective feature selection is vital for optimizing ML algorithm performance in complex datasets.

Purpose of the Study:

• To evaluate the efficacy of Support Vector Machine (SVM) for diagnosing Chronic Kidney Disease (CKD).
• To compare the performance of wrapper and filter feature selection methods in reducing CKD dataset dimensions.
• To identify the optimal feature selection technique for enhancing SVM accuracy in CKD diagnosis.

Main Methods:

• Utilized the Support Vector Machine (SVM) classification algorithm for CKD diagnosis.
• Applied two primary feature selection approaches: wrapper and filter methods.
• Within the wrapper approach, employed classifier subset evaluator with greedy stepwise and wrapper subset evaluator with Best First search.
• Within the filter approach, used correlation feature selection with greedy stepwise and filtered subset evaluator with Best First search.

Main Results:

• The Support Vector Machine (SVM) classifier demonstrated high diagnostic accuracy for CKD.
• The filtered subset evaluator with the Best First search engine achieved the highest accuracy rate of 98.5%.
• This specific feature selection method outperformed other evaluated wrapper and filter approaches.

Conclusions:

• The combination of Support Vector Machine (SVM) and the filtered subset evaluator with Best First search offers a highly accurate method for Chronic Kidney Disease (CKD) diagnosis.
• Feature selection plays a critical role in optimizing ML model performance for medical applications.
• This approach holds promise for improving early detection and management of CKD.