Optimization assisted ensemble classification for prediction of chronic kidney disease

Chronic kidney disease (CKD), with its growing prevalence, high risk of progressing to end-stage renal disease, and poor prognosis for morbidity and mortality, poses a substantial challenge to the healthcare system. It is gradually evolving into a global health emergency. The main causes of this condition are low water intake and unhealthy eating habits. A person can only function without their kidneys for an average of 18 days before they require dialysis or a kidney transplant. It is crucial to have accurate methods for CKD early prediction. For that reason, we have developed a CKD prediction model, which includes the following three working stages. In the pre-processing stage, the input image quality is enhanced by the use of anisotropic filtering. Along with the conventional Local Ternary Pattern (LTP), as well as Statistical features we have introduced an Improved Local Directional Pattern (ILDP) to provide better outcomes. In the final stage, we developed an ensemble model with classifiers such as Improved Convolutional Neural Network, Long short-term memory (LSTM), and Bi-Directional GRU (Bi-GRU) for classification. For optimal tuning of neural network parameters, we have developed an algorithm named Combined Coot Jaya Optimization (CCJO). Also, the result proves that this proposed CCJO-based ensemble model can offer superior outcomes in the CKD prediction task.