Rice Datathon '24 - Chevron

Inspiration

We are driven by the use of data science to solve industry problems.

What it does

We use an XGBoost model to predict Oil Peak Rate

How we built it

We internally regressed relevant features like number of stages and included mean imputation and overall preprocessing over well characteristics and fluid usage.

Challenges we ran into

There was not enough number of stages data so we created an internal learner, a linear regression based on available data, to predict this value and use it as a reference.

Accomplishments that we're proud of

Predicted number of stages matches the statistical characteristics of the original variable distribution. We also identified numerous georeferenced patterns and performed well geoclustering as part of the feature engineering process, which points at some useful next steps for researching on the problem.