oMLt

• 

  Hyperparameters (PosWeight + NN) vs Accuracy

• 

  Screenshot of keywords from losing submissions

Inspiration

We were inspired by our own difficulties breaking into the Hackathon scene as freshmen. We wanted to create a product which could level the playing field for future generations of hackers.

What it does

oMLt uses a KNN machine learning algorithm on training sets of former hackathon submissions to predict, through some clever linear algebra, whether a particular submission is likely to win.

How we built it

The project is made up of four parts: a scraper running on python, a parser on ruby, a transforming program which turns project descriptions and tags into linear data, and the testing suites themselves. We divided work up by our various strengths, with Camilo focusing on the scraping, Anand on the ruby components, and Kevin on the architecture of the transforming program.

Challenges we ran into

This was our first out-of-class experience implementing a machine learning algorithm. Therefore, we had to spend lots of time dealing with nitty gritty linear algebra formulas as well as confusing python syntax. Getting the 3D plots to run was especially difficult.

Accomplishments that we're proud of

We're most proud of our ability to form one cohesive product out of many tools and languages. Our project was one long pipe between web data and an ML prediction, and bridging that gap required the successful interplay of many moving parts. Were it not for the collaboration on the part of all our team members, that mission would have been impossible.

What we learned

We learned that you can still have fun while creating a successful product :)

What's next for oMLt

oMLt will only get bigger, just you wait ;)

Built With

• machine-learning
• python
• ruby