An Empirical Analysis of Machine Learning Classifiers for Clinical Decision Making in Asthma

An empirical analysis of the various machine learning classifiers on Asthma data and their performance assessment are presented. As an integral part of preprocessing the data, feature reduction and outlier detection are carried out, following which classifiers are applied to the same to distinguish asthmatic subjects from those of non-asthmatic ones. In cases where the classifiers perform poorly, their performance is boosted via adaptive classifiers. Feature ranking is adopted to reduce the number of features that are considered for the final process of classification while support vector machine with radial basis function is deployed to eliminate outliers. The performance is evaluated via cross validation and random sampling with variable training sizes. The various performance metrics evaluated for the system including precision, recall, F1-measure and classification accuracy indicate good results. The study can be used as a guideline for the effective application of knowledge discovery techniques on clinical data in order to explore hidden knowledge representations.