Data Envelopment Analysis with Missing Data an Expectation Maximization Approach

Şenel T., Terzi Y., Gümüştekin S., Cengiz M. A.

PONTE, cilt.72, sa.3, ss.36-43, 2016 (AHCI)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 72 Sayı: 3
  • Basım Tarihi: 2016
  • Doi Numarası: 10.21506/j.ponte.2016.3.27
  • Dergi Adı: PONTE
  • Derginin Tarandığı İndeksler: Arts and Humanities Citation Index (AHCI)
  • Sayfa Sayıları: ss.36-43
  • Ondokuz Mayıs Üniversitesi Adresli: Evet

Özet

In data envelopment analysis (DEA), the input and output data must be complete for the

comparison of decision-making units (DMU). Some data may be missing due to various

reasons. In such cases, examination of the DMUs with missing data is excluded. Excluding of

the examination of some decision-making units may change the effectiveness of other

decision-making units. To estimate missing observations as closely to the actual value as

possible may be more useful than excluding of the examination of decision-making units. In

this study, a problem with the complete input and output data was selected from the DEA

literature, and then some data were randomly deleted and turned into a problem with missing

data. After the missing data had been estimated with the Expectation Maximization (EM)

Algorithm, problems with missing and complete data were examined using DEA and the

results were compared.