Stock selection using data envelopment analysis-discriminant analysis

Abstract

Discriminant Analysis (DA) is a statistical method that can predict the category of a newly sampled observation. Some investigations involving Data Envelopment Analysis (DEA) have conducted discriminant analysis using DEA, an approach called DEA-DA. Base on both the DEA-DA method of Sueyoshi (2004) and Sueyoshi and Hwang (2004) and the concepts of investment decision, this study measures from the financial indices of Taiwanese banks to construct a discriminant function that allow investors to distinguish between superior and inferior stocks in terms of stock returns for the upcoming year. Analytical results demonstrate that a 100% hit rate for DEA-DA and leave-one-out cross validation with DEA-DA attains an accuracy rate of 85% confirming that DEA-DA is a highly effective method of classifying stocks into groups to help investors availably for investors to select stocks.

keywords:

• Data envelopment analysis
• discriminant analysis
• cross validation