Predicting Kidney Failure using an Ensemble Machine Learning Model: A Comparative Study

Within the domain of healthcare, medical diagnoses invariably necessitate comprehensive physical examinations and meticulous reviews of patients' medical histories. This study presents a pioneering approach that amalgamates distinct models to augment the effectiveness of our proposed model, resulting in an enhanced Pearson correlation for streamlined feature selection. Our hybrid model integrates a blend of chosen models. The Meta classifier employed is the Random Forest classifier. The primary aim of this experiment is to evaluate prominent machine learning classification techniques, thereby identifying the most precise classifier. Our investigation critically examines the existing machine learning model, with a specific emphasis on accuracy. For practical implementation, we incorporated four high-performing ML models and developed a unique hybrid model. This hybrid model, named "Hybrid," is built to optimize predictive capabilities using the UCI Chronic Kidney Failure dataset. In terms of accuracy, our results indicate that the Decision Tree classifier achieves approximately 97% accuracy, the Random Forest achieves 98% accuracy, Gaussian Naïve Bayes achieves about 93% accuracy, Naïve Bayes attains nearly 99% accuracy, while our proposed hybrid model achieves the highest accuracy of 100% and outperforms all other models on the same dataset. Prominent machine learning models considered for predicting the onset of Chronic Kidney Failure (CKF) encompass Naïve Bayes, Random Forest, Decision Tree, Support Vector Machine, K-nearest neighbor, Linear Discriminant Analysis, Gradient Boosting, and neural networks. This study contributes to advancing predictive accuracy in healthcare diagnosis using a novel hybrid model.