Thesis Type: Postgraduate
Institution Of The Thesis: Istanbul Commerce University, Fen Bilimleri Enstitüsü, Turkey
Approval Date: 2017
Thesis Language: Turkish
Student: YASEMİN BAHAR YÜCEL
Supervisor: BAĞDATLI KALKAN SEDA
Abstract:Classification and regression trees (CART) are techniques which don't require hypothesis. CART is named "classification tree" when the dependent variable is categorical, and it is named "regression tree" when the dependent variable is continuous. CART shows the model of the significant relations between dependent and independent variables as trees. Since it is easy to interpret, can be implemented to large data sets, and doesn't require hypothesis; it is often used recently. The most important feaures of Classification and Regression Tree (CART) and Chi-square Automed Interaction Detection (CHAID) algorithms are to be able to include continuous and categorical datas at the same time, and to be able to show the independent variables that are effective on the dependent variables on a tree diagram which is very easy to understand. The aim of this study is to determine the factors that affect the happiness levels of people living in Turkey with classification and regression trees. It is aimed to determine these factors, by using CART and CHAID algorithms on both classification and regression trees. Besides, trees are composed by using different beginning and test datas for classification and regression trees; the differencies between them are examined and the results are interpreted comparatively.