11DEA‐DISCRIMINANT ANALYSIS

11.1 INTRODUCTION

This chapter discusses discriminant analysis (DA)1. DA2 is a classification method that can predict the group membership of a newly sampled observation. In the use of DA, a group of observations, whose memberships are already identified, are used for the estimation of weights (or parameters)3 of a discriminant function by some criterion such as minimization of misclassifications or maximization of correct classifications. A new sample is classified into one of several groups by the DA results.

In the last two decades, Sueyoshi (1999a, 2001b, 2004, 2005a, b, 2006), Sueyoshi and Kirihara (1998), Sueyoshi and Hwang (2004a), Sueyoshi and Goto (2009a,b,c, 2012c, 2013b)4 and Goto (2012) proposed a new type of non‐parametric DA approach that provides a set of weights of a linear discriminant function(s), consequently yielding an evaluation score(s) for determining group membership. The new non‐parametric DA is referred to as “DEA‐DA” because it maintains discriminant capabilities by incorporating the non‐parametric feature of DEA into DA. An important feature of DEA‐DA5 is that it can be considered as an extension of frontier analysis, using the L1‐regression analysis discussed in Chapter 3. Moreover, a series of research efforts has attempted to formulate DEA‐DA in the manner that they can handle a data set that contains zero and/or negative values. Thus, the development of DEA‐DA in this chapter is different from that of DEA discussed in previous ...

Get Environmental Assessment on Energy and Sustainability by Data Envelopment Analysis now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.