U11 is a recommendation system for Graduate Schools, targetted towards international students applying to colleges in the United States. The application is a HTML site, with a Flask python backend system.

Work by: Anirban Chatterjee, Ravish Chawla, Tanmay Rajpurohit, Yanchao Feng, Yannick Schröcker

The codebase is structured as follows:

• data/
  • maindata.tsv - main dataset used for classification/training
  • trained_models - trained models built by SVM learner
• raw-data/ - raw data from scraper
• scraper/ - scraper and cleaning python scripts
  • scraper.py - scraper for website 1point3acres.com
  • cleaning/
    • json_to.tsv.py - convert json data to a tsv file
    • transform_scores.py - split GRE/TOEFEL grades based on distribution
    • merge_scores.py - fill GPA scores for missing records by user
    • split_university.py - split rows with multiple universities
    • fix_unicode.py - fix unicode values lost by openrefine
• site/ - main website
  • index.html - main HTML website
  • script.js - JS script to control site
  • styles.css - CSS stylings for site
  • app.py - FLASK api
  • create_list.py - classifies data onto training models
  • training/ - model building/training
    • classification.py - script that builds training models
• images/ - images of the application

The following python libraries are required for the application to run properly:

• Python (version 3)
  • flask
  • flask-cors
  • pandas
  • numpy
  • scipy
  • sklearn

These should be installed via pip, corresponding to version 3 of Python.

Start be ensuring that the following port is not being used by another application:

netstat -tn | grep 5000

If netstat is not installed, it can be installed via apt-get, homebrew, or choco. The above command should not print any output. If it prints a connection that is using the port on localhost, kill the application before continuing.

The application runs with a Backend and the Frontend. Having the above dependencies installed, start the backend service as follows:

cd site/
python app.py

This will start the python API server on http://localhost:5000

Start a new service (while keeping the above service running):

cd site/
python -m http.server

The application will be viewable now. It will be accesible via a browser window, on http://localhost:8000.

Upon opening the website, follow the on-screen instructions to go through the different steps, and view the results of the app. Images of what different steps of the application are can be viewed in the project sub-directory images/.

http://graduateranking.herokuapp.com/main.htm