An Intelligent Clinical Support System For The Early Diagnosis Of The Chronic Kidney Disease

Chronic kidney disease is a fatal renal illness, affecting a huge percentage of the world's population. This disease is life-threatening, possessing high morbidity and mortality rates. This disease being asymptomatic even at higher stages, its early diagnosis is of vital importance. Due to improvements in the branch of biosciences, Electronic Health Records have created vast volumes of data. With the help of this data, machines can be trained to produce precise predictions for the efficient and accurate diagnosis of this condition. For such a scenario, the machine learning-based approach plays a critical role. These methods assist practitioners in making early, accurate, and unambiguous clinical decisions about this renal disease. In this work, an efficient clinical diagnosis system is developed using Support Vector Machine and Logistic Regression algorithms. The cross-validation and testing results for each algorithm have been calculated using features chosen using the chi-squared feature selection approach. The data set for the problem is obtained from the UCI repository. Moreover, this work is focused on accurate as well as cost-efficient prediction of CKD by identification of the minimal set of relevant features along with some machine learning techniques like hyperparameter tuning to increase the model’s accuracy in the terms of disease identification. The SVM model proposed in this system reported accuracy of 99.17%. The usefulness of such a system lies in the mitigation of human loss that can be achieved by an accurate diagnosis of this renal ailment.