Machine Learning in Healthcare Analytics: A State-of-the-Art Review

Abstract

The use of machine learning (ML) models have become a crucial factor in the growing field of healthcare, ushering in a new era of medical research and diagnosis. This study rigorously reviews research publications published in reputable journals during the last five years. The pace and dynamic nature of machine learning in the healthcare domains demonstrated by the arduous criteria, which are used to sort through these articles. Disease-centric analysis uncovered a wide range of deep learning and machine learning models which are designed to address particular medical problems. Convolutional neural networks (CNNs), one of the most complex deep learning architectures, coexist with more conventional statistical models like logistic regression and support vector machines. CNNs are particularly prominent when it comes to disorders that need picture processing, which highlights the significant influence of deep learning in deciphering complex medical patterns. The popularity of ensemble methods, such as Random Forest, Gradient Boosting, and AdaBoost, indicates that their ability to combine predictive capability and strengthen model resilience is well acknowledged. Hybrid techniques, which integrate the advantages of many models, provide novel approaches to tackle distinct healthcare problems. This research also sheds light on a nuanced approach for model selection, wherein deep learning models performs well with huge datasets and image analysis, while statistical and ensemble models provides better results with numerical and categorical data. The adaptability needed in healthcare analytics is shown by hybrid models, which frequently combine standard models for classification with deep learning for feature extraction. The present review can endow problems related to ML in healthcare domain, possible solutions, potential directions and some knowledge to the researchers working in this field.