Enhancing Chronic Kidney Disease Diagnosis With FPA-DNN

Chronic kidney disease (CKD) is a critical global health concern, requiring accurate and early diagnosis to prevent its progression. Deep Learning (DL)models have shown promise in automating CKD classification from patient data, but the challenge lies in optimizing these models for peak performance. In this study, we propose an advanced approach for CKD classification, introducing the Flower Pollination Algorithm (FPA) for hyperparameter tuning of a Deep Neural Network model. To further enhance model performance, we employ an algorithm known as Oppositional Crow Search (OCS) which is utilized in feature selection and fine-tuning. The integration of FPA and OCS not only optimizes the DNN architecture but also refines the dataset for improved feature relevance. Experimental results demonstrate that the FPA-DNN model, guided by OCS, outperforms other conventional approaches in terms of accuracy, sensitivity, specificity, and Fl-score. This research showcases the potential of metaheuristic algorithms in improving DL models for CKD classification and opens avenues for enhanced medical diagnostics through artificial intelligence techniques. The results from the simulations demonstrated high exceptional efficacy for FPA-DNN approach.