Prediction of Chronic Kidney Disease Using Deep Neural Networks

Prediction of Chronic Kidney Disease Using Deep Neural Networks

T Sai Sushanth1, T Sanjay Siddarda2, A Sathvika3, A Shruthi4, A Sreelekha5,

Thayyaba Khatoon6, Atcha Kalyani7, Student12345, Professor6, Assistant Professor7

Department of Artificial Intelligence and Machine Learning (AI $ ML) Malla Reddy University, Maisammaguda, Hyderabad

2111CS020452@mallareddyuniversity.ac.in1 , 2111CS020484@mallareddyuniversity.ac.in2

2111CS020500@mallareddyuniversity.ac.in3 , 2111CS020518@mallareddyuniversity.ac.in4

, 2111CS020539@mallareddyuniversity.ac.in5 , thayyaba.khatoon16@gmail.com6, , Kalyani.a@mallareddyuniversity.ac.in7

Keywords:

Chronic kidney disease (CKD)

Deep Neural Network (DNN) Machine learning

Mortality risk

Medical research

Predictive modelling

A B S T R A C T

Diabetes and hypertension stand as the foremost culprits behind Chronic Kidney Disease (CKD), a condition characterized by a gradual decline in renal function over time. Researchers worldwide rely on Glomerular Filtration Rate (GFR) and markers of kidney damage to pinpoint the onset and progression of CKD. Individuals afflicted with CKD face heightened mortality risks, underscoring the urgency of early detection and intervention by healthcare professionals. However, diagnosing the diverse ailments associated with CKD poses a formidable challenge. This paper introduces an innovative deep learning model tailored for the early identification and prediction of CKD. The primary aim is to develop a deep neural network and assess its efficacy against other contemporary machine learning approaches. The model learns complex patterns and relationships within the data to predict the likelihood of CKD onset or progression. We evaluated the proposed framework using a large-scale dataset of patients with diverse demographic and clinical characteristics. Our results demonstrate superior performance compared to traditional prediction models, achieving high accuracy, sensitivity, and specificity in identifying individuals at risk of CKD. Notably, the proposed model achieved a superior performance with DNN reaching an accuracy of 96% by demonstrating its effectiveness in detecting attacks on IoT network Additionally, we conducted feature importance analysis to elucidate the factors contributing most significantly to the predictive outcome. Overall, our paper underscores the potential of deep neural network approaches in enhancing CKD risk layer and personalized patient care, thereby facilitating early intervention and improved clinical outcomes.