التنبؤ بمرض الكلى املزمن باستخدام التعلم اآللي

Abstract

The objective of this work is to develop an intelligent system for predicting chronic kidney disease based on machine learning techniques, by studying the impact of preprocessing stages and feature selection on the performance of different models. Various data preprocessing techniques were applied, including missing value imputation, outlier handling, and data scaling, in addition to feature selection techniques such as RFE, MI, and Chi-Square. Several algorithms were also compared, namely Decision Tree, Random Forest, Extra Trees, XGBoost, and Stacking Ensemble, using different evaluation metrics as well as execution time. The results showed that ensemble learning models achieved better performance compared to traditional models, where the Stacking Ensemble model recorded the highest accuracy, while the Random Forest model achieved an effective balance between high accuracy and execution speed. The study also showed that using five selected features contributed to improving performance and reducing model complexity. Based on the obtained results, the Random Forest model was adopted as the final model and integrated into an interactive web application using Flask, allowing direct and fast prediction results through an easy and efficient user interface.