Inspiration

Rumor has it, if you're good at basketball, you can easily get into CIT. Penn admits many outgoing girls. UChicago likes admitting artists. MIT is excited to see science olympiad recipients. Some people spend thousands of dollars hiring professional agencies to help with college application. Others post "Chance me" posts desperately on College Confidential. How do colleges decide who to admit? We are all curious, but no one actually knows.

What it does

AdmitMe has its unique machine learning algorithm to generate your chances of getting into top US colleges. Students can either log in through their common application account, or manually fill in a form of their personal information. Our web app will provide you a list of colleges of your choice that you are likely to be accepted (in order). You can check detailed statistic information on all the properties. We also provide you with personalized guidance on how you can most efficiently improve your chances, through improving parts of your application and the way you describe your extracurricular activities.

How we built it

We aggregated over 8000 detailed application information for the top 25 US colleges. The data comes from scraping "Chance" thread in College Confidential and a lot of data processing. The model is trained through diverse standard exam scores, senior course load, extracurricular activities and personal background. We used MongoDB as our database for all the data needed. The web app uses Python, Django, Javascript and html.

Challenges we ran into

Only one person on the team is familiar with the Django framework, which made it difficult to chain everything together. We ran into compatibility issues with Python 2.7 and Python 3 on the last day of the Hackathon, having to modify the whole algorithm. Natural language processing of self-reported data is challenging as well.

Accomplishments that we're proud of

We're unique. None of the other college counseling apps are using detailed natural language processing and machine leaning like us. We're able to process 8000 valid and detailed student data from nothing. We developed a 70% accuracy algorithm on top college prediction. We were also able to put a full stack website together in less than 36 hours.

What we learned

  • Django is difficult
  • Processing raw data to a good dataset is really time consuming. If we can find better data source in the future, that's a much easier way
  • Collaboration is very important for full stack programming, especially using Django

What's next for AdmitMe

  • We will develop a business plan to make our product more complete. (It's more of a prototype right now) The scalability needs to be improved.
  • Talk to colleges and agencies to aggregate more complete data to develop a more accurate algorithm
  • Market for our product, we are considering B to B (selling to agencies) or B to C (directly to students). We realized there is a huge market in China as well, where 3 out of our 4 team members are from.
Share this project:
×

Updates