Equipment Failure Prediction

Inspiration

Challenge by ConocoPhilips

What it does

Takes in sensor data and uses a neural network to determine if there will be an equipment failure or not.

How I built it

I used sklearn and pandas to clean up the csv data input. Then I oversampled the failure cases to account for class imbalance. I then made a binary classifier in Keras and fit it to the data and used it to predict the test data for failures which were then evaluated on Kaggle using the F1 metric.

Challenges I ran into

I ran into a lot of confusion on optimizing for F1 and trying to do that directly. There was also a lot of challenge in handling missing data and interpolating and filling null values.

Accomplishments that I'm proud of

I am proud of my first non classroom data science project and my first neural network (in or out of class).

What I learned

I learned about SMOTE oversampling and made my first Keras neural network.

What's next for Equipment Failure Prediction

Better handling of null values More hyperparameter tuning

Built With

• keras
• numpy
• pandas
• python
• scikit-learn