Research Information System
Thesis Type: Doctorate
Institution Of The Thesis: Gazi Üniversitesi, Fen Bilimleri Enstitüsü, Turkey
Approval Date: 2016
Student: MUSTAFA İSA DOĞAN
Supervisor: HACI HASAN ÖRKCÜ
Open Archive Collection: AVESIS Open Access Collection
Abstract:Classification problems are popular problems which are often encountered in data mining, statistics, economy and industry. In this study, new classification models, which can be used to solve the problems of two groups and multi-group classification, have been developed. The proposed new model for the two groups of classification model is the mixture of the Pendharkar and Troutt (2014) model that is based on the Data Envelopment Analysis BCC model and the two-stage classification model proposed by Sueyoshi (2004). The proposed new approach is examined in detail on an example taken from Pendharkar and Troutt (2014) and it is also observed that the classification performance of the proposed method is better than the two other methods in the simulation study. The developed multi-group classification model is a new classification model that is based on mathematical programming and radial basis neural networks. In multi-group mathematical programming classification models proposed in the literature, the data is assumed to have positive values. In the first stage of the new proposed model, after the data sets containing negative values are moved to positive data set space with the help of radial basis neural network, classification score for each unit, in a similar manner to Satapaty et al. (2009), is estimated with the help of a linear regression equation generated for each unit. In the second stage, as in Lam and Moy (1996), classification of units is made by mathematical programming model that makes threshold test. In the simulation study, it is observed that the proposed method performs better than many other multi-group classification methods which can run on negative data