A predictive model for the early prognosis and characterization of asthma

Abstract

An attempt to attain a good balance between optimal sensitivity and specificity of the predictive models in the case of sparse categorical data has been made in this paper by proposing a hybrid decision support system that integrates unsupervised and supervised learning methods at two different stages to explore the advantages of both. The system handles the categorical data without any encoding procedures involved. However, it requires one to use numerical categorical data in the place of labeled categorical data. The responses recorded in ACQ’s are largely numerical data characterizing the presence or absence with 1 and 0 respectively. A single optional value indicative of a third response is often used representing an unknown response. The primary aim of the proposed work is to provide a platform for efficient outcome prediction that can assist in shared decision making. Shared decision making in the context of healthcare relates to a strategy where both the clinicians and patient themselves decide on par with the management of the disease.