Chronic Kidney Disease Detection using Machine Learning Algorithms: A Sri Lankan Study

Chronic Kidney Disease (CKD) is quickly becoming the world’s most dangerous disease, and it induces other diseases. The diagnosis and treatment of CKD in its early stages are very important since patients can correct intermittent kidney function to halt the progression of this disease at this stage. And also, at this early stage, there are no obvious signs, and the cost of diagnosing and treating CKD is relatively significant, especially in developing countries like Sri Lanka. The CKD has had a severe impact on the residents of Anuradhapura district in Sri Lanka; the origin and exact aetiology are not yet fully understood. Recently, early CKD detection has been significantly supported by machine learning techniques. In this study, we developed a CKD detective system to diagnose chronic kidney diseases at any stage using efficient feature selection methods and machine learning based on the data collected from Anuradhapura district, Sri Lanka. Data pre-processing was performed, especially filling in missing values using the KNN imputation method with different K values. The correlation matrix, the variance inflation factor, and the wrapper feature selection methods are used to select the best sets of features in the CKD dataset for all K values. Four supervised machine learning algorithms were utilized to create the most accurate machine learning models. The Random Forest classification achieved the greatest accuracy (96%) of these machine learning techniques for the different K values. The proposed system could be implemented to detect CKD, especially in Sri Lanka, and ensure that the health and quality of life of the people are improved.