Statistical Classification Using Selection and Ranking Methodologies with Statistical Learning

The subject of Statistical Classification is concerned with identifying and allocating future observations into one of the pre-categorized classes based on the characteristics of the objects. Typically, these decisions to classify and categorize the objects have been dependent on identifying a system of classification, and from there, determining attributes for sorting. In past decades, from discriminant analysis, various methods have been developed for classification. In particular, the rise of artificial intelligence (AI), machine learning, and statistical learning theory has made it possible to consider improving the existing methods along with new developments and more comprehensive schemes in conjunction with data-driven methods. In this dissertation, we propose innovation using multiple decision-theoretic perspectives, such as the Indifference-Zone (IZ) approach and the Subset Selection (SS) method, to improve and clarify how these classification decisions can be made.