Skill Prediction in GitHub
I worked on this project along with two other classmates. Our primary aim was to use the GitHub data dump provided by the MSR Challenge 2014 and produce a tool or a finding that could contribute to the Software Engineering community. Our decision to build a skill prediction model for GitHub had multiple motivations. Firstly, unlike StackOverflow, GitHub does not have a formal model for projecting developer/participant reputation in terms of their skills. A skill prediction model can be used as a basis for deciding developer reputation. Secondly, given that GitHub is a social coding platform where developers collaborate on communal goals, a skill prediction model based on the actual work of a developer can help individuals develop trust and locate reputable contributors in the community. Thirdly, the manifestation of this model as a tool can be used by developers to build their resume through their work and recruiters to locate potential candidates to pursue for interviews. Here is the project paper for more details.
Learning and Challenges
This project was highly open ended and the most challenging yet rewarding aspect of the project was the phase of ideation. All three of us involved in the project contributed immensely towards brainstorming an idea that was novel, workable and marketable. We conducted multiple ideation sessions and argued on the pros and cons of each idea before coming to a conclusion. In this process, we learned a lot from each other. Another challenge was the time limit. We were provided two weeks of time to come up with the code along with a written report to present our project. While I learned execution under time pressure, another interesting experience was writing the project paper. It gave me a lot of insight into how research projects are represented through words.
MongoDB Java API