Inspiration

We were inspired by global issues with water quality, including the problem of eutrophication. This affected on of our teammates, as they had family in Flint, Michigan during the Flint water crisis.

What it does

This algorithm is build to provide accurate classifications of water samples based on 9 different characteristics. This allows consumers and water monitors to classify water as drinkable or not.

How we built it

We build this algorithm using Sci-kit and a Support Vector Classification model.

Challenges we ran into

  1. The dataset we were using had many missing values, so we used several solutions and had to pick the one with consistently the best accuracy.
  2. We switched from algorithm to algorithm as well, going from neural network in tf, to decision trees, and finally to SVC.

Accomplishments that we're proud of

We were able to achieve ~70% accuracy and implement it into a functioning website in just 24 hours.

What we learned

We learned tremendously about machine learning, as well as the importance of using technology to address contemporary issues.

What's next for Water Quality

Share this project:

Updates