Özyeğin University, Çekmeköy Campus Nişantepe District, Orman Street, 34794 Çekmeköy - İSTANBUL
Phone : +90 (216) 564 90 00
Fax : +90 (216) 564 99 99
E-mail: info@ozyegin.edu.tr
![Quick Link](https://www.ozyegin.edu.tr/sites/default/files/upload/quick-link.png)
Thesis Defense - Elif Ersoy (MSIE)
Elif Ersoy - M.Sc. Industrial Engineering
Asst.Prof. Erinç Albey – Advisor
Asst.Prof. Enis Kayış – Co-Advisor
Date: 26.08.2020
Time: 14:00
Location: This meeting will be held ONLINE. Please send an e-mail to gizem.bakir@ozyegin.edu.tr in order to participate in this defense.
A Genetic Algorithm for Constructing High Accuracy Decision Trees
Thesis Committee:
Asst.Prof. Erinç Albey, Özyeğin University
Asst.Prof. Enis Kayış , Özyeğin University
Asst.Prof. Mehmet Önal, Özyeğin University
Assoc.Prof. Mehmet Güray Güler, Yıldız Technical University
Asst.Prof. Yasin Ulukuş, İstanbul Technical University
Abstract:
Decision trees are one of the most widely used classification methods because of their ease of implementation and explainable nature. Conventional decision tree algorithms construct trees by using myopic, greedy top-down induction strategies. CART (classification and regression tree), ID4, C4.5 are well-known examples of such algorithms. Yet, there are disadvantages of these greedy, myopic methods. One major disadvantage is that while determining next split , they do not consider the possible impact of this decision on future splits. In other words, their myopic nature impedes attaining global optimal solution. In the literature, mathematical programming (i.e., mixed integer programming (MIP) models) is also employed to find optimal trees. Solving constructed models using solvers that guarantee global optimal solution is possible yet remains intractable for even medium size instances due to NP-Hard nature of the problem. This study seeks to construct high accuracy trees than that of the conventional tree construction algorithms, such as CART; and to attain a near-optimal solutions in shorter time than MIP models. To achieve these, we propose a genetic algorithm with a genuine chromosome structure. We also address the selection of the initial population by considering a blend of randomly generated solutions, solutions from the CART, and solutions from the mathematical model, which are constructed for reduced problem instances. We test the performance of the proposed genetic algorithm using five different datasets, with varying bounds on the depth of the resulting trees and using different initial population blends within the mentioned varieties. Results reveal that the performance of the proposed genetic algorithm is superior to that of CART in almost all datasets used in the analysis.
Bio:
Elif ERSOY graduated from Beşiktaş Atatürk Anatolian High School in 2013. She received her B.S. degree in Industrial Engineering from Özyeğin University in June 2017. After graduation, she joined Master of Science program in Industrial Engineering at Özyeğin University and has been working under the supervision of Asst. Prof. Enis Kayış and Asst. Prof. Erinç Albey. Her research mainly focuses on data science.
----------------------------
Kişisel Verilerin Korunması/Aydınlatma Metni
Özyeğin Üniversitesi Fen Bilimleri Enstitüsü bünyesindeki programlarda, izleyici olarak katılınan online platformlarda gerçekleştirilen aleni akademik faaliyetlerin sadece idari ve yasal yükümlülüklerin yerine getirilmesi amacıyla kayıt altına alındığı, söz konusu kayıt ile elde edilen kişisel verilerin ilgililer tarafından başka bir amaçla kullanılamayacağı hususunda şahsınızı aydınlatır, toplantıya gönüllü katılım ile bu hususu beyan ve taahhüt etmiş olduğunuzu belirtir, gereğini önemle rica ederiz.
Fen Bilimleri Enstitüsü