The goal of this project was to explore, understand, analyze, and generate insights from a dataset of BCycle rides from 2014-2018. To do this, our approach was to understand the data we were working with, correct any data for potential errors, and iterate as we explored through various questions we posed to ourselves. A few of those topics we examined are:
1) Membership Segments - How can we classify the various memberships into distinct categories? - How is the ride activity between those segments different? 2) Kiosk - Which kiosks are used the most and least? Why might that be? - How does demand vary for different kiosks? 3) Change over Time - Can we visualize how bikes are used throughout the day? - How does seasonality affect usage? 4) Predictions - Can we forecast future BCycle demand? - Can we model effects like seasonality, day of week effects, and holidays on future demand?
A few of the results we obtained:
1) Better understood consumer behavior around memberships and ride data which allowed us to build profiles for various user segments 2) Understood which data was anomylous likely due to technical errors (e.g. 0 min ride duration) 3) Generated heatmaps of kiosk utilization over time 4) Developed a demand forecast model that modeled time series effects including seasonality, holidays, and day of week effects 5) Identified a growth pattern in recent BCycle ridership usage 6) Developed ideas for business implications on customer targeting and future growth opportunities