Tuning Multi-Layer Perceptron by Hybridized Arithmetic Optimization Algorithm for Healthcare 4.0

Abstract

Healthcare 4.0 has been enabled by the recent advances in several different fields, including the artificial intelligence (AI) and Internet of things (IoT), among others. It aims to support the doctors in making decisions as fast as possible by providing early prediction of the illness, that could save lives of the patients. Typical workflow includes IoT sensors that collect the data of the patients, while the AI is employed to analyze this data by utilizing a variety of machine learning approaches to predict and classify the illness. This manuscript aims to suggest a healthcare framework that consists of a neural network that is optimized by a hybridized arithmetic optimization algorithm. In the proposed framework, metaheuristics is used to simultaneously perform the network training and tune the hyper-parameters of the multi-layer perceptron neural network. The performance of this framework has been evaluated on three healthcare benchmark datasets, with the task to classify the fatal conditions including heart diseases, breast cancer and diabetes. The obtained results were compared to five other cutting-edge metaheuristics that were employed in the same framework with the same task. The proposed hybridized arithmetic optimization algorithm achieved superior level of performance, indicating that this structure could assist the medical workers in early diagnostics.