Predicting the Kidney Diseases by Using Machine Learning Techniques

• Published in: ITM Web of Conferences Vol. 57; p. 1011
• Main Authors: Sreenivasa, N., Pawaar, Sudesh, Sparsh, Shaurya, Ramesh Naidu, P.
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Les Ulis EDP Sciences 2023
• Subjects: Algorithms; chronic kidney disease; Datasets; decision tree; Decision trees; Kidney diseases; logistic regression; Machine learning; Missing data; Regression; Regression analysis; Support vector machines; support vectors machines
• ISSN: 2271-2097; 2431-7578; 2271-2097
• DOI: 10.1051/itmconf/20235701011

CKD (Chronic Kidney Diseases) is a persistent medical state categorized by the kidney damage that hinders their ability to effectively filter blood. Over time, this progressive disease can result in kidney failure. This project compares the performance of the Support Vectos Machines (SVM), logistic regression and Decision Tree algorithms for predicting the risk of CKD. In this project, the dataset utilized comprises a total of 25 attributes, consisting of 11 numerical features and 14 nominal features. In the training of machine learning algorithms for prediction, all 400 instances from the dataset are utilized. Among these instances, 250 are labeled as CKD cases, indicating the presence of chronic kidney disease, while the remaining 150 instances are categorized as non-CKD cases, denoting the absence of the condition. We utilized the UCI dataset, which underwent preprocessing to handle missing data. Using Python, we trained and built Support Vectors Machines (SVM), Logistic Regression, and Decision Tree models. The accuracy achieved with SVM was 97.3%, Logistic Regression was 93.8%, and Decision Tree yielded 95%, which are notable results.