MULTI-MODEL STACK ENSEMBLE DEEP LEARNING APPROACH FOR MULTI-DISEASE PREDICTION IN HEALTHCARE APPLICATION

In the modern era of computers, numerous disciplines are witnessing the development of massive data volumes. Statistics are important in healthcare engineering because they provide insights into various diseases and match patient data. These datasets serve two functions: improving illness prediction and examining large data reservoirs to identify previously unknown disease patterns. Leveraging deep learning models, it becomes feasible to detect and forecast the early stages of numerous diseases based on individual health conditions. Nonetheless, the current landscape of illness prediction encounters several challenges such as inadequate large-scale datasets, logistical delays, the imperative for more precise and dependable predictions, and the intricacy of the models themselves. This paper introduces an innovative method for disease prediction utilizing deep learning, particularly focusing on an ensemble-based multi-disease prediction model. Datasets for lung cancer, cervical cancer, chronic renal disease, Parkinson's illness, and HCC survival are sourced from the trustworthy UCI repository for experimentation. A robust stacked deep ensemble model is proposed combining the InceptionResNetV2, EfficientNetV2L, and Xception architectures. This model integrates pre-processing techniques and employs the Tuna Swarm Optimization (TSO) Algorithm for feature selection in executing multi-label disease prediction. The suggested deep learning algorithms' performance is evaluated using criteria such as precision, specificity, sensitivity, accuracy, and error rate. This assessment demonstrates the potential of the recommended method to significantly contribute to the healthcare system by offering consistent and reliable predictions across various types of illnesses, as shown in a comparative analysis against existing models.