During few hours of the day, certain bike checkpoints remain empty. Nearby users will have to find their way to another station to use a bike. If someone is paying for a membership pass to use these bikes, they would become frustrated if there weren’t any nearby.
What it does
We identified and looked at, which times were the busiest for B-Cycle usage, As well as which kiosks are most trafficked, and what their peak times were. Given a specific time and kiosk, our project will estimate how many users will be needing B-Cycles, so that management can prepare to not have an empty rack.
How we built it
We imported Azure Auto ML, which tried to fit a variety of different models to our input data (i.e. kiosk ID, hour of the day).
Challenges we ran into
Azure Auto ML took longer than we expected to run (so far 1), and we could not interrupt it as we need it to produce our final model to make predictions with.
Accomplishments that we're proud of
Teamwork, Our graphs/visualization were pretty organized and well-developed.
What we learned
How to graph for specific time frames, sorting hours from timestamps, turning datasets into dictionaries, analyzing trends so that we know what data to ignore, learning methods for classifying data.
What's next for B-Cycle Kiosk Demand
Tuning hyper-parameters of the best model output from Auto ML, running more tests, and using them to improve them to improve our accuracy