Great ideas can come from the smallest thought. And machine learning can definitely count as a small though!
What it does
A machine learning model that will predict asset failure for a given date or weather condition input, to be used by the rail industries itself.
How we built it
Using a Random Forest Model, Decision tree, Neural Networks.
Challenges we ran into
Overfitting (near perfect predictions for non perfect data).
Accomplishments that we're proud of
Getting a sound proof of concept. Finding papers to back machine learning models that work.
What we learned
Railways are prone to weather related damage and fault finding can be done using weather data and simple logistic regression models (from past research)
What's next for Scottish Data
Include data from weather forecasts to give a preventative measure