International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
Renal Vision Of Chronic Kidney Disease With The Aid Of Machine Learning N.Deepika(Asst prof) Department of CSE Kakatiya Institute of Technology & Science Warangal,India
Thanishka Boora Department of CSE Kakatiya Institute of Technology & Science Warangal,India
Deekshitha Chitti Department of CSE Kakatiya Institute of Technology & Science Warangal,India
Sunayana Deepthi Shamanthula Department of CSE Kakatiya Institute of Technology & Science Warangal,India
Pavan Teja Reddy Pingili Department of CSE Kakatiya Institute of Technology & Science Warangal,India
------------------------------------------------------------------------***----------------------------------------------------------------------Abstract—A major worldwide health issue, chronic kidney disease (CKD) requires precise prediction models for early detection and treatment. In order to predict CKD, this study investigates the efficacy of a number of ml approaches, such as Random Forest (RF), Support Vector Machines (SVM),kNearest Neighbors (k-NN), and Gated Recurrent Units (GRU). In addition, the SHapley Additive exPlanations (SHAP) feature reduction technique is used to improve the interpretability of the model and pinpoint important predictors. The temporal dependency-capturing GRU model is compared to RF, SVM, and k-NN, and critical variables influencing each model's prediction of CKD are identified and prioritized using SHAP analysis. The comparative performance of these algorithms in terms of sensitivity, specificity,accuracy, is demonstrated by the results, which offer insightful information for the creation of CKD prediction models that are comprehensible and accurate, with potential uses in early detection and better patient outcomes. Keywords— Random Forest (RF), Chronic Kidney Disease(CKD), Support Vector Machines (SVM), Gated Recurrent Units (GRU), k-Nearest Neighbors (k-NN), SHapley Additive exPlanations (SHAP)
I. INTRODUCTION A major global health issue that affects millions of people and provides enormous problems to healthcare systems is chronic kidney disease[1], or CKD.You may encounter a number of health issues as a result of CKD.[2,3,4] In order to carry out prompt interventions and individualized treatment programs, early detection and precise prediction of CKD progression are essential. Machine learning models have become extremely efficient tools for predictive analytics in the medical industry in recent years [5]. Leveraging largescale numeric datasets, these models can uncover complex relationships within the data and contribute to more precise prognostic outcomes. In this context, our study focuses on
© 2024, IRJET
|
Impact Factor value: 8.226
|
advancing Chronuc kidney disease (CKD) [6] prediction methodologies through the effective integration of diverse machine learning models, namely Gated Recurrent Unit (GRU)[7] Random Forest (RF)[8], Logistic Regression (LR)[9,10] andSupport Vector Machine (SVM)[9]. The goal is to use the distinct characteristics of each model to increase prediction accuracy and to improve interpretability by utilizing the SHapley Additive exPlanations (SHAP) feature reduction method. The utilization of a robust dataset plays a important role in the success of predictive modeling for CKD. Our study underscores the significance of a comprehensive and well-curated numeric dataset, enriched with diverse patient attributes and clinical parameters. By harnessing the wealth of information encapsulated in these datasets, our models aim to discern intricate patterns and relationships, allowing for a more nuanced understanding of CKD progression factors. Furthermore, we explore the potential of the Gated Recurrent Unit (GRU) [7] to capture temporal dependencies, Random Forest (RF)[8] for handling non-linear relationships, Support Vector Machine (SVM) [9] for discerning complex decision boundaries, and Logistic Regression (LR) [9,10] for its simplicity and interpretability. The amalgamation of these models is facilitated by the SHAP feature reduction method, enhancing both the accuracy and interpretability of the CKD prediction models. In our pursuit of accurate and interpretable CKD prediction, the [11,12,13,14,15,16] SHapley Additive exPlanations (SHAP) feature reduction method takes center stage. By quantifying the effect of single attributes on model predictions, SHAP provides valuable insights into the driving factors behind [1] CKD progression. This method not only enhances the transparency of our models but also empowers healthcare practitioners to make informed
ISO 9001:2008 Certified Journal
|
Page 2453