Predictor Cal Poly Senior Project

The goal of this project is given a list of available developers and open issues in a Github repository to predict which issue is best suited for each developer to work on next.

This project accomplishes this by data-mining the previous contributions by each developer and finding patterns in what they other issues they have succesfully completed. The idea is that someone who has worked on similar features in the past will complete the new issues more quickly and with fewer defects.

Techniques

This project uses a few different techniques to determine similarity between two texts:

1. Transform each text into a set of words and find how many words are shared between the texts
2. The Levenshtein edit distance between the texts
3. Cosine similarity between the raw texts
4. Cosine similarity between the texts after they have been stemmed and tokenized

Validation

The Statistics page steps through time to make a guess with only the data available at the time that the issue was available to work on. It outputs the amount of correct guesses that it made and can give an indication of confidence for predicting the current dataset.

Data Issues

Not all projects remeber to include the issue number in each pull request so many projects are missing significant amounts of data that could help build the prediction model.

Getting Started

This project uses vagrant for testing and developing

Change the secret key in predictor.wsgi Change the MySQL password in the vagrant/provision-app.sh script Change the GitHub user and password in the views.py file Change the MySQL password in the Database.py file

Go to the vagrant folder and run: ./prepare-vagrant.sh vagrant up

After the virtual machine is up: vagrant ssh Get the machines IP from ifconfig Change the IP in /etc/apache2/sites-available/Predictor.conf from 127.0.0.1 to your IP sudo service apache2 restart

Congrats, you are running the predictor

Usage

Import Project

Under this tab, you can import a project by its full Github name. IE: bbeck13/RhythmRunner This step takes a long time for large projects

Analyze Project

Under this tab, you can pick from available developers, and available issues and it will produce in order the most relevent issues for the developer to work on.

Generate CSV Data

This tab will produce a CSV file of the confidence ratings from each prediction model for each issue that has already been completed.

Statistics

This tab will produce the number of correct predictions the predictor has made on the project

Delete Project

Delete the current project

Switch Project

Switch between already imported projects