%0 Journal Article
%T Efficacious End User Measures¡ªPart 1: Relative Class Size and End User Problem Domains
%A E. Earl Eiland
%A Lorie M. Liebrock
%J Advances in Artificial Intelligence
%D 2013
%I Hindawi Publishing Corporation
%R 10.1155/2013/427958
%X Biological and medical endeavors are beginning to realize the benefits of artificial intelligence and machine learning. However, classification, prediction, and diagnostic (CPD) errors can cause significant losses, even loss of life. Hence, end users are best served when they have performance information relevant to their needs, this paper¡¯s focus. Relative class size (rCS) is commonly recognized as a confounding factor in CPD evaluation. Unfortunately, rCS-invariant measures are not easily mapped to end user conditions. We determine a cause of rCS invariance, joint probability table (JPT) normalization. JPT normalization means that more end user efficacious measures can be used without sacrificing invariance. An important revelation is that without data normalization, the Matthews correlation coefficient (MCC) and information coefficient (IC) are not relative class size invariants; this is a potential source of confusion, as we found not all reports using MCC or IC normalize their data. We derive MCC rCS-invariant expression. JPT normalization can be extended to allow JPT rCS to be set to any desired value (JPT tuning). This makes sensitivity analysis feasible, a benefit to both applied researchers and practitioners (end users). We apply our findings to two published CPD studies to illustrate how end users benefit. 1. Introduction Biological compounds and systems can be complex, making them difficult to analyze and challenging to understand. This has slowed applying biological and medical advances in the field. Recently, artificial intelligence and machine learning, being particularly effective classification, prediction and diagnostic (CPD) tools, have sped applied research and product development. CPD can be described as the act of comparing observations to models, then deciding whether or not the observations fit the model. Based on some predetermined criterion or criteria, a decision is made regarding class membership ( or ). In many domains, class affiliation is not the end result, rather it is used to determine subsequent activities. Examples include medical diagnoses, bioinformatics, intrusion detection, information retrieval, and patent classification. The list is virtually endless. Incorrect CPD output can lead to frustration, financial loss, and even death; correct CPD output is important. Hence, a number of CPD algorithms have been developed and the field continues to be active. Characterizing CPD effectiveness, then, is necessary. For example, CPD tool developers need to know how their particular modification affects CPD performance, and
%U http://www.hindawi.com/journals/aai/2013/427958/