Pregler Mann - ConocoPhillips downhole equipment failure

Inspiration

I had experience working with rare event classification type problems in my coursework this summer in an optional class. I am glad I took it because as soon as I heard the problem, I knew I wanted to try to apply what I had learned.

What it does

It takes a rare event classification problem, balances the dataset on the rare events to make the model capable of accurately predicting the holdout set.

How we built it

We used basic python and sklearn.

Challenges we ran into

Run times in Kaggle and sleep deprivation.

Accomplishments that we're proud of

Getting a submission! This is both of our first times using Kaggle and first times at a datathon.

What we learned

That if we build off of our strengths, that we can come up with interesting solutions to complicated problems.

What's next for Pregler Mann - ConocoPhillips downhole equipment failure

Hopefully we win and ConocoPhillips can recognize Seth and myself as some of the only people who truly understood the problem in a way that produced a valid model.

Built With

• python
• sklearn