GitHub Developer Profiler Based on Latent Dirichlet Allocation

Tez Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Florida State University, Amerika Birleşik Devletleri

Tez Danışmanı: Sonia Haiduc

Tezin Onay Tarihi: 2014

Tezin Dili: İngilizce

Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu

Özet:

It is frequently happening event that a developer leaves and a new developer is hired or assignment of a developer is changed, in the software companies that have many employees. After that, some role assignments are happened and newcomers or developers, who are assigned to the project that they have not worked on, might have trouble with catching how the project goes on. In these cases, they need to ask help from the more experienced colleagues, but also they need to know expertise of the colleagues to get a useful answer.

On the other hand, pair programming is an agile software development technique in which two programmers work together and nowadays, it is used very common in software companies. The project manager is responsible for deciding pairs. Consistency of assigned pairs is one of the most important factors that affect efficiency. The project manager needs to know developers expertise deeply to make the best decision for the company.

In this project, I aim to implement such a tool which creates developers’ profiles using source codes and commit messages of each developer in project repositories. I am looking at repositories to obtain history of the project which means to have far more information, instead of looking at final version of the project. To illustrate, while final version of the study case, Imagej[1], has 136 thousand lines of source code, its repository has more than 6 millions lines of information.