ConocoPhillips offered the challenge to make predictions on their sensors to reduce cost of replacing parts. We decided to take them on that challenge to learn about machine learning.
What it does
Our python code generates a new maximum accuracy of .9933 through our neural network.
How we built it
Challenges we ran into
Largest obstacle was getting past accuracy of .9908 along with deciding how to clean the data.
Accomplishments that we're proud of
Maximum accuracy within .162% of first place with use of Neural Networks.
What we learned
We learned how to use Tensorflow, Adaboost classifier, we learned how to compare and contrast different machine learning models, process of converting Python files to HTML files, and learn the advantages and disadvantages of different optimizations.
What's next for Sensor Failure Prediction with Neural Networks.
Continue to increase accuracy and test other models.