PhD Researcher2015–2018
CDOT Construction Duration Predictor
Neural network that predicts highway construction durations for the Colorado DOT — packaged as a web app now used statewide.
- Keras
- TensorFlow
- Python
- Django
Built an artificial neural network for the Colorado Department of Transportation to predict construction duration for highway projects based on estimated material quantities and geographic attributes.
The ANN proved substantially faster and more reliable than CDOT's existing methods, and significantly more accurate than the linear regression models other academic institutions had developed.
CDOT was pleased enough with the results to commission a web app wrapping the model — now used by civil engineers across the state on every CDOT project. Funded by Transportation Pooled Fund TPF-5(260); other state DOTs in the study group later asked us to extend the model to their states.