Past Research Projects:

Reduction of Industrial Data Requirements: It is a huge effort for companies to collect data. For some companies data collection can be out of the question, because they cannot afford the required time or the resources.  For some other companies that already have data collection processes, a reduction in data collection effort without sacrificing from the performance can mean hundreds of thousands of dollars. In this project we are working on methods that can aid data collection. For example a company that wishes to use the data coming from another company for its own needs may find our paper on cross-company data acquisition an interesting read. Similarly, a manager thinking how to reduce the number of projects for which s/he needs to collect data can refer to another paper of ours, where we propose a method to identify the most informative software projects.

Stability Issues in Software Effort Estimation: The process of estimating the work effort required to complete a software project can be defined as SEE. Regardless of the decades of research effort spent on SEE, we are still unable to see stable conclusions. On the contrary, different studies indicate different conclusions, i.e. we lack stability in the results of SEE studies. Our research identified the likely culprits of the instability as: 1) The SEE data sets, 2) how the estimation methods are used, 3) error measures used to evaluate the estimation methods and 4) the sampling methods used for the evaluation of the methods. This project is an empirical investigation of the factors leading the instability in SEE and it aims at providing explanations and cures for these factors.

Inductive Engineering: Academic and industrial data mining are based on very different practices: e.g. the academic data mining focuses on algorithms, whereas industrial data mining focuses on users. Although these differences are partially described by individual case studies in SE, they have never been put in an organized and clear format. This project intends to provide guidelines for the knowledge transfer between academic and industrial data mining. For our paper on the inductive engineering manifesto you can refer here or see our Facebook page here. In this project I have the privilege to work with Dr. Tim Menzies, Dr. Thomas Zimmermann and Dr. Christian Bird; who are also responsible for the lion's share of the work.

This was an industry funded research project to design and implement a software effort prediction model for IBTech, which is the software development subsidiary of an international bank in Turkey. This project involves developing a data analytics model, which will predict the development effort required for each phase of the software development life-cycle of a new software project. Ultimate aim of this project is to assist project managers while making resources allocations. I was fortunate enough to work with Dr. Ayse Bener, Dr. Ayse Tosun Misirli and Dr. Bora Caglayan in this project.

This was an industry funded research project to develop a measurement and defect prediction model for Turkcell, which is the biggest GSM provider in Turkey. The project involves mining software metrics repositories of Turkcell to use software process data, churn metrics and static code attributes for the purposes of defect prediction and product analysis.