Implementation of Discretisation and Correlation-based Feature Selection to Optimize Support Vector Machine in Diagnosis of Chronic Kidney Disease

This study aims to improve the accuracy of the classification algorithm for diagnosing chronic kidney disease. There are several models of data mining. In classification, the Support Vector Machine (SVM) algorithm is widely used by researchers worldwide. The data used is a chronic kidney disease dataset taken from the UCI machine learning repository. This data consists of 25 attributes and 11 numeric data attributes, and 14 negative attributes. To call continuously, discrete data is used. Meanwhile, data is selected using Correlation-based Feature Selection (CFS) to reduce irrelevant and redundant data. The research results by applying discretization and feature selection based on correlation for classification in the SVM algorithm with 10-fold cross-validation show an increase in accuracy of 0.5%. The classification of the vector machine support algorithm in the diagnosis of chronic kidney disease produces an accuracy of 99.25%, and after applying discretization and correlation-based feature selection, produces an accuracy of 99.75%. Implementation of discretion and correlation-based feature selection to optimize support vector machine for diagnosis of chronic kidney disease has increased accuracy by 0.5%. The proposed method is feasible as a method of diagnosing chronic kidney disease.