Machine learning classification approach for asthma prediction models in children

Background Machine Learning refers to a methodology in the domain of data analytic that automates the systematic building of the model. It permits the discovery of unseen insights from an enormous datasets by means of suitable methods which involve repetitive learning gathered from data devoid of being programmed explicitly. The aim of this work is to explore machine learning strategies that are able to compensate with the weaknesses of existent asthma development predictive models in children. The objective of this review is to identify, compare, and summarize the existing machine and deep learning classification models for asthma prediction in children. Methodology A substantial number of asthma development prediction models in children, such as conventional methods of risk factors, logical regression, and the hybrid of both statistical methods and risk factors existed. This study was performed following the guideline of Preferred Reporting Items for systematic Review and Meta Analysis (PRISMA). We carried a search for relevant studies from 2011-2021 using various online databases such as Google Scholar, Science Direct and PubMed on 23 July, 2021 to extract relevant papers on asthma prediction Models in children using machine learning and deep learning approaches. Result The weaknesses associated with these existent asthma development predictive models in children include: they cannot be used as an appropriate tool for the implementation of decision support in electronic medical records, reduced clinical impact as well as low predictive accuracy. It was observed that ANN and SVM were among the best-performing algorithms in some machine learning comparative asthma prediction in children. Conclusion This work concludes that there is a gradual increase of machine and deep learning algorithms for asthma prediction in children and that these approaches have shown greater predictive performance in pediatric asthma than the conventional existing models.